Fraud Detection in Python - solutions.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"Fraud Detection in Python - solutions.ipynb","provenance":[],"collapsed_sections":[]},"kernelspec":{"name":"python3","display_name":"Python 3"}},"cells":[{"cell_type":"markdown","metadata":{"id":"6fwBZkxc2gW6","colab_type":"text"},"source":["[datacamp link](https://www.datacamp.com/courses/fraud-detection-in-python)\n","\n","Useful online tutorial:\n","\n","[Credit Card Fraud Detection\n","](https://www.kaggle.com/mlg-ulb/creditcardfraud/kernels)\n"]},{"cell_type":"code","metadata":{"id":"ufbYb6vynIMK","colab_type":"code","colab":{}},"source":["import matplotlib.pyplot as plt\n","import numpy as np\n","import pandas as pd\n","import seaborn as sn"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"rNNFBD1NpYXh","colab_type":"code","colab":{}},"source":["from sklearn.model_selection import train_test_split, GridSearchCV\n","from sklearn.linear_model import LogisticRegression\n","from sklearn.metrics import accuracy_score, average_precision_score\n","from sklearn.metrics import precision_recall_curve\n","from sklearn.metrics.cluster.supervised import homogeneity_score\n","from sklearn.metrics.cluster.unsupervised import silhouette_score"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"Hj3Ed9RisrWG","colab_type":"code","outputId":"61e322b7-3fd7-4b9c-9574-0b7e9725c27b","executionInfo":{"status":"ok","timestamp":1573284237084,"user_tz":-480,"elapsed":895,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":71}},"source":["from imblearn.over_sampling import SMOTE\n","from imblearn.pipeline import Pipeline \n","from sklearn.ensemble import RandomForestClassifier, VotingClassifier\n","from sklearn.preprocessing import MinMaxScaler\n","from sklearn.cluster import MiniBatchKMeans, DBSCAN\n","from sklearn.metrics import classification_report, confusion_matrix, roc_auc_score\n","from nltk.corpus import stopwords \n","from nltk.stem.wordnet import WordNetLemmatizer\n","import gensim\n","from gensim import corpora"],"execution_count":0,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/externals/six.py:31: DeprecationWarning: The module is deprecated in version 0.21 and will be removed in version 0.23 since we've dropped support for Python 2.7. Please rely on the official version of six (https://pypi.org/project/six/).\n","  \"(https://pypi.org/project/six/).\", DeprecationWarning)\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"ppNDx_jHl8iS","colab_type":"text"},"source":["**Course Description**\n","\n","A typical organization loses an estimated 5% of its yearly revenue to fraud. In this course, you will learn how to fight fraud by using data. For example, you'll learn how to apply supervised learning algorithms to detect fraudulent behavior similar to past ones, as well as unsupervised learning methods to discover new types of fraud activities. Moreover, in fraud analytics you often deal with **highly imbalanced datasets** when classifying fraud versus non-fraud, and during this course you will pick up some techniques on how to deal with that. The course provides a mix of technical and theoretical insights and shows you hands-on how to practically implement fraud detection models. In addition, you will get tips and advice from real-life experience to help you prevent making common mistakes in fraud analytics."]},{"cell_type":"markdown","metadata":{"id":"r-m9m-mml-f5","colab_type":"text"},"source":["## 1. Introduction and preparing your data\n","\n","In this chapter, you''ll learn about the typical challenges associated with fraud detection, and will learn how to resample your data in a smart way, to tackle problems with imbalanced data."]},{"cell_type":"markdown","metadata":{"id":"ScX3ACZWnQek","colab_type":"text"},"source":["### Introduction to fraud detection\n","\n"]},{"cell_type":"markdown","metadata":{"id":"Gci8N1hLYpFC","colab_type":"text"},"source":["#### Checking the fraud to non-fraud ratio\n","In this chapter, you will work on `creditcard_data.csv`, a dataset containing credit card transactions data. Fraud occurrences are fortunately an extreme minority in these transactions.\n","\n","However, Machine Learning algorithms usually work best when the different classes contained in the dataset are more or less equally present. If there are few cases of fraud, then there's little data to learn how to identify them. This is known as **class imbalance**, and it's one of the **main challenges of fraud detection**."]},{"cell_type":"code","metadata":{"id":"tmrUgs-zmCyL","colab_type":"code","outputId":"2e844c64-14a7-4b96-bbe2-07bbfd7f846f","executionInfo":{"status":"ok","timestamp":1573273678781,"user_tz":-480,"elapsed":742,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":646}},"source":["# Import pandas and read csv\n","import pandas as pd\n","# df = pd.read_csv(\"creditcard_data.csv\")\n","df = pd.read_csv(\"creditcard_sampledata_3.csv\")\n","# Explore the features available in your dataframe\n","print(df.info())"],"execution_count":0,"outputs":[{"output_type":"stream","text":["<class 'pandas.core.frame.DataFrame'>\n","RangeIndex: 5050 entries, 0 to 5049\n","Data columns (total 31 columns):\n","Unnamed: 0    5050 non-null int64\n","V1            5050 non-null float64\n","V2            5050 non-null float64\n","V3            5050 non-null float64\n","V4            5050 non-null float64\n","V5            5050 non-null float64\n","V6            5050 non-null float64\n","V7            5050 non-null float64\n","V8            5050 non-null float64\n","V9            5050 non-null float64\n","V10           5050 non-null float64\n","V11           5050 non-null float64\n","V12           5050 non-null float64\n","V13           5050 non-null float64\n","V14           5050 non-null float64\n","V15           5050 non-null float64\n","V16           5050 non-null float64\n","V17           5050 non-null float64\n","V18           5050 non-null float64\n","V19           5050 non-null float64\n","V20           5050 non-null float64\n","V21           5050 non-null float64\n","V22           5050 non-null float64\n","V23           5050 non-null float64\n","V24           5050 non-null float64\n","V25           5050 non-null float64\n","V26           5050 non-null float64\n","V27           5050 non-null float64\n","V28           5050 non-null float64\n","Amount        5050 non-null float64\n","Class         5050 non-null int64\n","dtypes: float64(29), int64(2)\n","memory usage: 1.2 MB\n","None\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"J3moiqPO_tcV","colab_type":"code","outputId":"1bcd1b19-be26-492a-808e-aff3fc4c82c8","executionInfo":{"status":"ok","timestamp":1573273585053,"user_tz":-480,"elapsed":735,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":241}},"source":["df.head()"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>Unnamed: 0</th>\n","      <th>V1</th>\n","      <th>V2</th>\n","      <th>V3</th>\n","      <th>V4</th>\n","      <th>V5</th>\n","      <th>V6</th>\n","      <th>V7</th>\n","      <th>V8</th>\n","      <th>V9</th>\n","      <th>V10</th>\n","      <th>V11</th>\n","      <th>V12</th>\n","      <th>V13</th>\n","      <th>V14</th>\n","      <th>V15</th>\n","      <th>V16</th>\n","      <th>V17</th>\n","      <th>V18</th>\n","      <th>V19</th>\n","      <th>V20</th>\n","      <th>V21</th>\n","      <th>V22</th>\n","      <th>V23</th>\n","      <th>V24</th>\n","      <th>V25</th>\n","      <th>V26</th>\n","      <th>V27</th>\n","      <th>V28</th>\n","      <th>Amount</th>\n","      <th>Class</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>258647</td>\n","      <td>1.725265</td>\n","      <td>-1.337256</td>\n","      <td>-1.012687</td>\n","      <td>-0.361656</td>\n","      <td>-1.431611</td>\n","      <td>-1.098681</td>\n","      <td>-0.842274</td>\n","      <td>-0.026594</td>\n","      <td>-0.032409</td>\n","      <td>0.215113</td>\n","      <td>1.618952</td>\n","      <td>-0.654046</td>\n","      <td>-1.442665</td>\n","      <td>-1.546538</td>\n","      <td>-0.230008</td>\n","      <td>1.785539</td>\n","      <td>1.419793</td>\n","      <td>0.071666</td>\n","      <td>0.233031</td>\n","      <td>0.275911</td>\n","      <td>0.414524</td>\n","      <td>0.793434</td>\n","      <td>0.028887</td>\n","      <td>0.419421</td>\n","      <td>-0.367529</td>\n","      <td>-0.155634</td>\n","      <td>-0.015768</td>\n","      <td>0.010790</td>\n","      <td>189.00</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>69263</td>\n","      <td>0.683254</td>\n","      <td>-1.681875</td>\n","      <td>0.533349</td>\n","      <td>-0.326064</td>\n","      <td>-1.455603</td>\n","      <td>0.101832</td>\n","      <td>-0.520590</td>\n","      <td>0.114036</td>\n","      <td>-0.601760</td>\n","      <td>0.444011</td>\n","      <td>1.521570</td>\n","      <td>0.499202</td>\n","      <td>-0.127849</td>\n","      <td>-0.237253</td>\n","      <td>-0.752351</td>\n","      <td>0.667190</td>\n","      <td>0.724785</td>\n","      <td>-1.736615</td>\n","      <td>0.702088</td>\n","      <td>0.638186</td>\n","      <td>0.116898</td>\n","      <td>-0.304605</td>\n","      <td>-0.125547</td>\n","      <td>0.244848</td>\n","      <td>0.069163</td>\n","      <td>-0.460712</td>\n","      <td>-0.017068</td>\n","      <td>0.063542</td>\n","      <td>315.17</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>96552</td>\n","      <td>1.067973</td>\n","      <td>-0.656667</td>\n","      <td>1.029738</td>\n","      <td>0.253899</td>\n","      <td>-1.172715</td>\n","      <td>0.073232</td>\n","      <td>-0.745771</td>\n","      <td>0.249803</td>\n","      <td>1.383057</td>\n","      <td>-0.483771</td>\n","      <td>-0.782780</td>\n","      <td>0.005242</td>\n","      <td>-1.273288</td>\n","      <td>-0.269260</td>\n","      <td>0.091287</td>\n","      <td>-0.347973</td>\n","      <td>0.495328</td>\n","      <td>-0.925949</td>\n","      <td>0.099138</td>\n","      <td>-0.083859</td>\n","      <td>-0.189315</td>\n","      <td>-0.426743</td>\n","      <td>0.079539</td>\n","      <td>0.129692</td>\n","      <td>0.002778</td>\n","      <td>0.970498</td>\n","      <td>-0.035056</td>\n","      <td>0.017313</td>\n","      <td>59.98</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>281898</td>\n","      <td>0.119513</td>\n","      <td>0.729275</td>\n","      <td>-1.678879</td>\n","      <td>-1.551408</td>\n","      <td>3.128914</td>\n","      <td>3.210632</td>\n","      <td>0.356276</td>\n","      <td>0.920374</td>\n","      <td>-0.160589</td>\n","      <td>-0.801748</td>\n","      <td>0.137341</td>\n","      <td>-0.156740</td>\n","      <td>-0.429388</td>\n","      <td>-0.752392</td>\n","      <td>0.155272</td>\n","      <td>0.215068</td>\n","      <td>0.352222</td>\n","      <td>-0.376168</td>\n","      <td>-0.398920</td>\n","      <td>0.043715</td>\n","      <td>-0.335825</td>\n","      <td>-0.906171</td>\n","      <td>0.108350</td>\n","      <td>0.593062</td>\n","      <td>-0.424303</td>\n","      <td>0.164201</td>\n","      <td>0.245881</td>\n","      <td>0.071029</td>\n","      <td>0.89</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>86917</td>\n","      <td>1.271253</td>\n","      <td>0.275694</td>\n","      <td>0.159568</td>\n","      <td>1.003096</td>\n","      <td>-0.128535</td>\n","      <td>-0.608730</td>\n","      <td>0.088777</td>\n","      <td>-0.145336</td>\n","      <td>0.156047</td>\n","      <td>0.022707</td>\n","      <td>-0.963306</td>\n","      <td>-0.228074</td>\n","      <td>-0.324933</td>\n","      <td>0.390609</td>\n","      <td>1.065923</td>\n","      <td>0.285930</td>\n","      <td>-0.627072</td>\n","      <td>0.170175</td>\n","      <td>-0.215912</td>\n","      <td>-0.147394</td>\n","      <td>0.031958</td>\n","      <td>0.123503</td>\n","      <td>-0.174528</td>\n","      <td>-0.147535</td>\n","      <td>0.735909</td>\n","      <td>-0.262270</td>\n","      <td>0.015577</td>\n","      <td>0.015955</td>\n","      <td>6.53</td>\n","      <td>0</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["   Unnamed: 0        V1        V2        V3  ...       V27       V28  Amount  Class\n","0      258647  1.725265 -1.337256 -1.012687  ... -0.015768  0.010790  189.00      0\n","1       69263  0.683254 -1.681875  0.533349  ... -0.017068  0.063542  315.17      0\n","2       96552  1.067973 -0.656667  1.029738  ... -0.035056  0.017313   59.98      0\n","3      281898  0.119513  0.729275 -1.678879  ...  0.245881  0.071029    0.89      0\n","4       86917  1.271253  0.275694  0.159568  ...  0.015577  0.015955    6.53      0\n","\n","[5 rows x 31 columns]"]},"metadata":{"tags":[]},"execution_count":21}]},{"cell_type":"code","metadata":{"id":"voX25fRps7Qj","colab_type":"code","outputId":"35357baf-7f89-4fce-9bdc-2937a918df1d","executionInfo":{"status":"ok","timestamp":1573273587173,"user_tz":-480,"elapsed":709,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":119}},"source":["# Count the occurrences of fraud and no fraud and print them\n","occ = df['Class'].value_counts()\n","print(occ)\n","\n","# Print the ratio of fraud cases\n","print(occ / len(df.index))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0    5000\n","1      50\n","Name: Class, dtype: int64\n","0    0.990099\n","1    0.009901\n","Name: Class, dtype: float64\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"9DxibVO-Zoo-","colab_type":"text"},"source":["####Plotting your data\n","\n","From the previous exercise we know that the ratio of fraud to non-fraud observations is very low. You can do something about that, for example by re-sampling our data, which is explained in the next video."]},{"cell_type":"code","metadata":{"id":"U74CQ3Jbalqb","colab_type":"code","outputId":"95194513-1756-4509-b5ff-63e7302fc6a7","executionInfo":{"status":"ok","timestamp":1573273693458,"user_tz":-480,"elapsed":745,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# match \n","df = df.drop(['Unnamed: 0'], axis=1) #\n","df.shape"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(5050, 30)"]},"metadata":{"tags":[]},"execution_count":26}]},{"cell_type":"code","metadata":{"id":"JUWYHeb-_ijz","colab_type":"code","colab":{}},"source":["# pre-loaded\n","def prep_data(df):\n","    X = df.iloc[:, 1:29]\n","    X = np.array(X).astype(np.float)\n","    y = df.iloc[:, 29]\n","    y=np.array(y).astype(np.float)\n","    return X,y"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"DT5QKnHMtIsQ","colab_type":"code","outputId":"342847b3-c5a5-4a6b-a022-2bfdff4d1d93","executionInfo":{"status":"ok","timestamp":1573273697708,"user_tz":-480,"elapsed":1620,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":267}},"source":["# Define a function to create a scatter plot of our data and labels\n","def plot_data(X, y):\n","\tplt.scatter(X[y == 0, 0], X[y == 0, 1], label=\"Class #0\", alpha=0.5, linewidth=0.15)\n","\tplt.scatter(X[y == 1, 0], X[y == 1, 1], label=\"Class #1\", alpha=0.5, linewidth=0.15, c='r')\n","\tplt.legend()\n","\treturn plt.show()\n","\n","# Create X and y from our above defined function\n","X, y = prep_data(df)\n","\n","# Plot our data by running our plot data function on X and y\n","plot_data(X, y)"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAXkAAAD6CAYAAABEUDf/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"Jl5TWcCXtLmV","colab_type":"text"},"source":["### Increasing successful detections using data resampling\n"]},{"cell_type":"markdown","metadata":{"id":"0fC2x40PtWEZ","colab_type":"text"},"source":["#### Applying SMOTE\n","\n","In this exercise, you're going to re-balance our data using the **Synthetic Minority Over-sampling Technique** (SMOTE). \n","\n","Unlike ROS, SMOTE does not create exact copies of observations, but creates new, synthetic, samples that are quite similar to the existing observations in the minority class. SMOTE is therefore slightly more sophisticated than just copying observations, so let's apply SMOTE to our credit card data. "]},{"cell_type":"code","metadata":{"id":"cfTKINbStMgV","colab_type":"code","colab":{}},"source":["from imblearn.over_sampling import SMOTE\n","\n","# Run the prep_data function\n","X, y = prep_data(df)\n","\n","# Define the resampling method\n","method = SMOTE(kind='regular')"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"w5rSrhrdE-Kr","colab_type":"code","colab":{}},"source":["# Create the resampled feature set\n","X_resampled, y_resampled = method.fit_sample(X, y)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"NpXkgPxfFH28","colab_type":"code","outputId":"e16a0540-98a1-48e8-9c3d-86835761f8d3","executionInfo":{"status":"ok","timestamp":1573273709982,"user_tz":-480,"elapsed":1233,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":267}},"source":["# Plot the resampled data\n","plot_data(X_resampled, y_resampled)"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAXkAAAD6CAYAAABEUDf/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"CSN8zlIIbAMI","colab_type":"text"},"source":["####Compare SMOTE to original data"]},{"cell_type":"code","metadata":{"id":"XIvVy8_WbD8u","colab_type":"code","colab":{}},"source":["# pre-loaded\n","def compare_plot(X,y,X_resampled,y_resampled, method):\n","    # Start a plot figure\n","    f, (ax1, ax2) = plt.subplots(1, 2)\n","    # sub-plot number 1, this is our normal data\n","    c0 = ax1.scatter(X[y == 0, 0], X[y == 0, 1], label=\"Class #0\",alpha=0.5)\n","    c1 = ax1.scatter(X[y == 1, 0], X[y == 1, 1], label=\"Class #1\",alpha=0.5, c='r')\n","    ax1.set_title('Original set')\n","    # sub-plot number 2, this is our oversampled data\n","    ax2.scatter(X_resampled[y_resampled == 0, 0], X_resampled[y_resampled == 0, 1], label=\"Class #0\", alpha=.5)\n","    ax2.scatter(X_resampled[y_resampled == 1, 0], X_resampled[y_resampled == 1, 1], label=\"Class #1\", alpha=.5,c='r')\n","    ax2.set_title(method)\n","    # some settings and ready to go\n","    plt.figlegend((c0, c1), ('Class #0', 'Class #1'), loc='lower center',\n","                  ncol=2, labelspacing=0.)\n","    #plt.tight_layout(pad=3)\n","    return plt.show()"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"Cqa72rqStcOv","colab_type":"code","outputId":"21598b74-65aa-428c-b089-27eb41afec7b","executionInfo":{"status":"ok","timestamp":1573273767141,"user_tz":-480,"elapsed":1531,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":397}},"source":["# Print the value_counts on the original labels y\n","print(pd.value_counts(pd.Series(y)))\n","\n","# Print the value_counts\n","print(pd.value_counts(pd.Series(y_resampled)))\n","\n","# Run compare_plot\n","compare_plot(X, y, X_resampled, y_resampled, method='SMOTE')"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0.0    5000\n","1.0      50\n","dtype: int64\n","1.0    5000\n","0.0    5000\n","dtype: int64\n"],"name":"stdout"},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAXkAAAEWCAYAAACDoeeyAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nXHZBgG2oxAAAAABJRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 2 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"P_BnXZ8-tq8D","colab_type":"text"},"source":["###Fraud detection algorithms in action\n"]},{"cell_type":"markdown","metadata":{"id":"-nTh0mSabOEg","colab_type":"text"},"source":["####Exploring the traditional way to catch fraud\n","\n","- First you'll define threshold values using common statistics, to split fraud and non-fraud. \n","\n","- Then, use those thresholds on your features to detect fraud. This is common practice within fraud analytics teams."]},{"cell_type":"code","metadata":{"id":"gcruNVbPtcRn","colab_type":"code","outputId":"70ede579-256e-49a7-f8ba-5f4bbd3fe754","executionInfo":{"status":"ok","timestamp":1573277365274,"user_tz":-480,"elapsed":784,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":85}},"source":["# Run a groupby command on our labels and obtain the mean for each feature\n","df.groupby('Class').mean()\n","\n","# Implement a rule for stating which cases are flagged as fraud\n","df['flag_as_fraud'] = np.where(np.logical_and(df['V1'] < -3, df['V3'] < -5), 1, 0)\n","\n","# Create a crosstab of flagged fraud cases versus the actual fraud cases\n","print(pd.crosstab(df.Class, df.flag_as_fraud, rownames=['Actual Fraud'], colnames=['Flagged Fraud']))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Flagged Fraud     0   1\n","Actual Fraud           \n","0              4984  16\n","1                28  22\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"W_qqhYX6hLjy","colab_type":"text"},"source":["####Using ML classification to catch fraud\n"]},{"cell_type":"code","metadata":{"id":"nytt_t82pPvS","colab_type":"code","colab":{}},"source":["from sklearn.model_selection import train_test_split\n","from sklearn.linear_model import LogisticRegression"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"Hc5Oac1duLvg","colab_type":"code","outputId":"513d4366-59c3-41cc-fdfa-fdcdc485b16e","executionInfo":{"status":"ok","timestamp":1573277489001,"user_tz":-480,"elapsed":918,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":292}},"source":["# Create the training and testing sets\n","X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)\n","\n","# Fit a logistic regression model to our data\n","model = LogisticRegression()\n","model.fit(X_train, y_train)\n","\n","# Obtain model predictions\n","predicted = model.predict(X_test)\n","\n","# Print the classifcation report and confusion matrix\n","print('Classification report:\\n', classification_report(y_test, predicted))\n","conf_mat = confusion_matrix(y_true=y_test, y_pred=predicted)\n","print('Confusion matrix:\\n', conf_mat)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Classification report:\n","               precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.89      0.80      0.84        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.94      0.90      0.92      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","Confusion matrix:\n"," [[1504    1]\n"," [   2    8]]\n"],"name":"stdout"},{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n","  FutureWarning)\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"EZ9P3K4FhScd","colab_type":"text"},"source":["####Logistic regression combined with SMOTE"]},{"cell_type":"code","metadata":{"id":"ltMX-E4py2Tk","colab_type":"code","colab":{}},"source":["# This is the pipeline module we need for this from imblearn\n","from imblearn.pipeline import Pipeline \n","\n","# Define which resampling method and which ML model to use in the pipeline\n","resampling = SMOTE(kind='borderline2')\n","model = LogisticRegression()\n","\n","# Define the pipeline, tell it to combine SMOTE with the Logistic Regression model\n","pipeline = Pipeline([('SMOTE', resampling), ('Logistic Regression', model)])"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"AgACEnREhWFW","colab_type":"text"},"source":["####Using a pipeline\n"]},{"cell_type":"code","metadata":{"id":"hXobnO3ruLyX","colab_type":"code","outputId":"8a57d4f9-203a-483e-c03f-8802d3919ff0","executionInfo":{"status":"ok","timestamp":1573277585813,"user_tz":-480,"elapsed":725,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":292}},"source":["# Split your data X and y, into a training and a test set and fit the pipeline onto the training data\n","X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)\n","\n","# Fit your pipeline onto your training set and obtain predictions by fitting the model onto the test data \n","pipeline.fit(X_train, y_train) \n","predicted = pipeline.predict(X_test)\n","\n","# Obtain the results from the classification report and confusion matrix \n","print('Classifcation report:\\n', classification_report(y_test, predicted))\n","conf_mat = confusion_matrix(y_true=y_test, y_pred=predicted)\n","print('Confusion matrix:\\n', conf_mat)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Classifcation report:\n","               precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.62      1.00      0.77        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.81      1.00      0.88      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","Confusion matrix:\n"," [[1499    6]\n"," [   0   10]]\n"],"name":"stdout"},{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n","  FutureWarning)\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"lchOovPFmC37","colab_type":"text"},"source":["## 2. Fraud detection using labelled data\n"," \n","Now that you're familiar with the main challenges of fraud detection, you're about to learn how to flag fraudulent transactions with supervised learning. You will use classifiers, adjust them and compare them to find the most efficient fraud detection model."]},{"cell_type":"markdown","metadata":{"id":"T_oV-ELXOhjC","colab_type":"text"},"source":["### Review of classification methods\n"]},{"cell_type":"markdown","metadata":{"id":"zoSD0zVmhexG","colab_type":"text"},"source":["####Natural hit rate\n","\n","First you need to explore how prevalent fraud is in the dataset, to understand what the \"**natural accuracy**\" is, if we were to predict everything as non-fraud. \n","\n","It's is important to understand which level of \"accuracy\" you need to \"beat\" in order to get a better prediction than by doing nothing. \n","\n","In the following exercises, you'll create our first random forest classifier for fraud detection. That will serve as the \"baseline\" model that you're going to try to improve in the upcoming exercises."]},{"cell_type":"code","metadata":{"id":"QG9jjmDhmH1L","colab_type":"code","outputId":"8e6ffc41-ad4b-4e04-b666-474daf464c51","executionInfo":{"status":"ok","timestamp":1573277639418,"user_tz":-480,"elapsed":752,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# Count the total number of observations from the length of y\n","total_obs = len(y)\n","\n","# Count the total number of non-fraudulent observations \n","non_fraud = [i for i in y if i == 0]\n","count_non_fraud = non_fraud.count(0)\n","\n","# Calculate the percentage of non fraud observations in the dataset\n","percentage = (float(count_non_fraud)/float(total_obs)) * 100\n","\n","# Print the percentage: this is our \"natural accuracy\" by doing nothing\n","print(percentage)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["99.00990099009901\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"kQtdRu0jh0Xx","colab_type":"text"},"source":["####Random Forest Classifier"]},{"cell_type":"code","metadata":{"id":"roeYHl2P79wp","colab_type":"code","colab":{}},"source":["# Import the random forest model from sklearn\n","from sklearn.ensemble import RandomForestClassifier\n","\n","# Split your data into training and test set\n","X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)\n","\n","# Define the model as the random forest\n","model = RandomForestClassifier(random_state=5)"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"vDGzpBApqFjU","colab_type":"text"},"source":["- Let's see how our Random Forest model performs without doing anything special to it. "]},{"cell_type":"code","metadata":{"id":"yrlj99lL791T","colab_type":"code","outputId":"6b976585-8b76-4d45-edfb-5bd5b602abaf","executionInfo":{"status":"ok","timestamp":1573277696801,"user_tz":-480,"elapsed":735,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":71}},"source":["# Fit the model to our training set\n","model.fit(X_train, y_train)\n","\n","# Obtain predictions from the test data \n","predicted = model.predict(X_test)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/ensemble/forest.py:245: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n","  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"],"name":"stderr"}]},{"cell_type":"code","metadata":{"id":"coLc6dNmqOzS","colab_type":"code","colab":{}},"source":["from sklearn.metrics import accuracy_score"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"oIfCFyBn8Eul","colab_type":"code","outputId":"ab3528d3-0bad-43a7-fc14-0904574eeb4a","executionInfo":{"status":"ok","timestamp":1573277726787,"user_tz":-480,"elapsed":649,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# Print the accuracy performance metric\n","print(accuracy_score(y_test, predicted))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0.9986798679867986\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"Ncn_-_DG8Qed","colab_type":"text"},"source":["**Random Forests vs. Decision Trees:**\n","\n","**Random Forest prevents overfitting** most of the time, by creating random subsets of the features and building smaller trees using these subsets. Afterwards, it combines the subtrees of subsamples of features, so it does not tend to overfit to your entire feature set the way \"deep\" Decisions Trees do."]},{"cell_type":"markdown","metadata":{"id":"sdGIYSaE8XZ5","colab_type":"text"},"source":["###Performance evaluation\n"]},{"cell_type":"markdown","metadata":{"id":"0k-kHmuBh_Lx","colab_type":"text"},"source":["####Performance metrics for the RF model\n","\n","With highly **imbalanced** fraud data, the **AUROC** curve is a more reliable performance metric, used to compare different classifiers. \n","\n","Moreover, the **classification** report tells you about the precision and recall of your model, whilst the **confusion matrix** actually shows how many fraud cases you can predict correctly. So let's get these performance metrics."]},{"cell_type":"code","metadata":{"id":"hHGYdgPQ8Eya","colab_type":"code","outputId":"8c6530f5-9a12-4c4d-8b21-06e36718fc93","executionInfo":{"status":"ok","timestamp":1573277766877,"user_tz":-480,"elapsed":759,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":221}},"source":["# Import the packages to get the different performance metrics\n","from sklearn.metrics import classification_report, confusion_matrix, roc_auc_score\n","\n","# Obtain the predictions from our random forest model \n","predicted = model.predict(X_test)\n","\n","# Predict probabilities\n","probs = model.predict_proba(X_test)\n","\n","# Print the ROC curve, classification report and confusion matrix\n","print(roc_auc_score(y_test, probs[:,1]))\n","print(classification_report(y_test, predicted))\n","print(confusion_matrix(y_test, predicted))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0.9993687707641197\n","              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.90      0.90      0.90        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.95      0.95      0.95      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","[[1504    1]\n"," [   1    9]]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"a_2_LeV2i3eR","colab_type":"text"},"source":["####Plotting the Precision Recall Curve\n"]},{"cell_type":"code","metadata":{"id":"0g-QLNSHIQdX","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":290},"outputId":"31bfb5d7-079d-4292-eece-2c8697a85ec1","executionInfo":{"status":"ok","timestamp":1573352764164,"user_tz":-480,"elapsed":699,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}}},"source":["import matplotlib.pyplot as plt\n","img = plt.imread('1_7J08ekAwupLBegeUI8muHA.png')\n","fig= plt.figure(figsize=(8,5))\n","plt.imshow(img)"],"execution_count":2,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7f6306f71898>"]},"metadata":{"tags":[]},"execution_count":2},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAeYAAAEACAYAAABmjtVQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niAKziIhIhfx/PfAk+rJ5IpYAAAAASUVORK5CYII=\n","text/plain":["<Figure size 576x360 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"WGjARmMHIsih","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":138},"outputId":"ea3b21ae-ca43-4959-aff4-f02693b0ad6f","executionInfo":{"status":"ok","timestamp":1573352881667,"user_tz":-480,"elapsed":667,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}}},"source":["import matplotlib.pyplot as plt\n","img = plt.imread('1_T6kVUKxG_Z4V5Fm1UXhEIw.png')\n","fig= plt.figure(figsize=(5,3))\n","plt.imshow(img)"],"execution_count":5,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7f6306de3f60>"]},"metadata":{"tags":[]},"execution_count":5},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAATkAAABoCAYAAACKRIcXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0\ndHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAASl0lEQVR4nO2dX4gkx3nAf18kyw+yIJIlhJCVWPKs\nEMclyKdDmBkhSMCJdg9y55dFeXDuITDzYMFuwAIF60EPyoMlbGYCwewYCzbBiViwvRbcDLETDCG3\nRNFekKXTibsZOzLWIUsygdhPcWJ/eejumZ7uqu6end750/f9oNnZ7prqr6q6vqmvvur6RFUxDMOo\nKr+1aAEMwzCOE1NyhmFUGlNyhmFUGlNyhmFUGlNyhmFUGlNyhmFUmpmUnIg8ISJXRWQoIs+UJZRh\nGEZZyFHXyYnITcA14LPAu8BrwJ+q6pXyxDMMw5iNWUZyjwJDVf2xqv4KeBk4W45YhmEY5TCLkrsX\n+Gns/3fDc4ZhGEvDzcd9AxFpAk2AW2+99ZGHHnrouG9pGMYNxqVLl36uqne5rs2i5K4D98X+/0R4\nbgJV7QJdgNOnT+vh4eEMtzQMw0gjIj/xXZvFXH0NWBOR+0XkFuBJ4JUZ8jMMwyidI4/kVPX/ROQp\n4B+Bm4CXVPWt0iQzDMMogZnm5FS1B/RKksUwDKN07I0HwzAqjSk5wzAqjSk5wzAqjSk5wzAqjSk5\nY4Xp0xJBpEV/oXkYy4wpOSMk6uzuo7WMGmB4jcsAzXOsLzKPuZDdPo3OcNECOhl2GmM5Gx2GBa+V\niqrO7XjkkUfUWGZ62gSFpvZG5wbarqOA1tuDBcpmuNtHddCuK6A0e95vLpRe0//8ZF2bAuBQPXrH\nRnLGmGhUUz/B2uhkja3dNnXgYO/C8f3aGvl4Rp21rWeDl8O7+0tpcg+vXQbg5IO1qa6VhSk5Y8zg\nCgdAffMMuY/csEMjMjVGtuyQTiMyoRq4Laik2VU0Xcykid/bN5fWbznMuti9iuThzMeRNrcuSjL3\no/Y5sZabdJIS6nwiWSuVptPym5yDKwdAHZfYWddKwzfEO47DzNXlJjJ7UlbPoK11UOptHUTpws+9\nZmQ+9bRJXduDyeuu/MemSWh+OdJOpAvvP2HSJGRK3yeQJTwTmNzJtBl56Khs8fqITPexuZhXF3n3\niOebZ7KlyzUS1GuullrnmXXiM5f998u+Nh1kmKum5IyQ6GFNdyKv8tPooa9rvT45T+RI6Jh7SSuN\n6H6udM3JRIHySGtkt0LrNdNpvXn4yzx9Xfg7clJBNnuRTA5F5qmrkYJzKYqS6zytMOPldtdJVh1n\nXpsSU3JGAdyT2tEDnPkr7eyU+Xm7H/Igbe4ve9iB02KNRxa5fceXR+4o0TXq8tRFXkeOrkeHd1QT\n5Z88fD8uZde5Jz8dK3fXM+Cvr+xr02JKzsgnGhGkjgwFVvSX2DGiGHkEPZ7CfHEzZJtQHP4Rpi+P\nI3VMj2LM7cgJJedNl6pD/8jbnX62OveXwzPCHInhlzGzDackS8mZ48EAxl6uenuQeEgusuXzQhSc\nCO/vdwE42F4bTVavbUN7oKjuTHgKg4noXGkJxD2J0ylX2+KiKr3A5ciGc+Lfn0ckg8vj573mcdpk\n5TXsNJDzsBvK2uwpu5x3OgbSXsgaZzbrwAF7F9LT/WXXubccwwvsHeBZZ5jVTjltWCKm5AxgyIW9\nwMu1eab4Exd1pGLu/3rYwQooT+o54oYda2KpS5r1HUUHwfKX7vMJz19uHg6P37DD812AJucSPdpd\nF332PekBalsX0Ytb1MIOf/naMDiXqptx+8Rlqp3ZzFnaU2Kdh2mSddJ/cdv/Q5dVxwXbsAxMyRnA\ngOCHeppf1eiX2N2Bi+bR6RRYWzHs04/3Yt9Sl2GHRnIZQ+0Mm67+O81ymZCoQzd7O4lRS1QXCSXg\nXHfoosbWReWif8jsbp/ag5wEOLjCoGAZjlznzjQd9i8Hlev8ocuq4yPU/1HJVXIicp+I/EBErojI\nWyKyFZ5/TkSui8jr4bFRllATr3v4jpyFR/1WgTVQ82RiXdaSvS7V3w+CcEz1qxp2vALfWT/XJGVW\nDTs0ZI29vWsT7ROlHcvWQtY22HjRV1HBerRWn6DjHGwzkbT/ItsHeZ0plgew/nSbOgdsxzIadhps\ndANzfiel1KO62MQ5ED75IDWCZ/JIr19522edc6FJ/nwi37LrfO1EwjTut5DzcG4TRsp92KHh7W/B\nGjz38551rQR8k3XRAdwDnAo/30YQUPoE8Bzwxbzvx4+ijoeRRy/r8E52J71QOUsbjh2fV6xIWeYh\nXtLhULC+pnT/jye987yIibSZXt3JievUPcjysrrzSJVvRgdM/Fk+UjMn28fj2HDVVbl1HlsPF08T\nr6dU/vE6Tj5XWdemhzK9q8B3gc/ORclN+VS4H/IFK7m89VyzdADDMFS1RO+qiHwS+DTwanjqKRF5\nQ0ReEpHbp8mrbIadBmvb4ZC73mYQuNYWT+jp05SNU2PrYo9Iyu7+MtithlE9Cis5EfkY8C1gW1V/\nAXwN+BTwMPAe8BXP95oicigihx9++GEJIrsJXNxNeqqhx2oViOZUgMvX7OV3wzgGCik5EfkIgYL7\npqp+G0BV31fVX6vqb4CvA4+6vquqXVU9raqn77rLGeC6FNZ30ut/KkfMeeGewB6/YL0UDg3DWAKK\neFcF+Abwtqp+NXb+nliyz0HgLTemIVp6wMgDl0lti9124LI/2D6fWjDab20EXrhmz+EBNIwbkyIj\nuQbweeAPE8tFXhCRN0XkDeAPgL84TkHnyXj5yWzLWHKJFkQCzYKLzWpbuwR6bnKJA/0WG8E6A9pP\nm4YzjIhcJaeq/6qqoqq/r6oPh0dPVT+vqr8Xnv8TVX2vdOm6G6u1HfeURItLqbcprpdqbD0bTuR1\nN8J66NMKNBz19m7GqnbDuPG4edECLCPrO4ruHO89osWlUKe9O6WjZH2HXrPLRhe6Gy1OtC+Hi0Xb\n7BbUcMEshGEsL8HKkNlZ7te6mj3vmruVnnPqt0bLXY468lrfiZafdNneDt5rnFpZGsYNgI3k5s2w\nQyM0LWn2Mt5XzGOdp9t1utHawOazUynLsn4lDWPZWe6R3II4NsfDsENjLZyHa/YcC4Sny+v8dux9\nw9H8nGEYcUzJzY0+rbWxo2Ewk709pHM+3OKmPWAQLivpbizJZgSGsUSYknMQLCwucBRWVH1asjFy\nDgxmfCNj2DlPMIhr8uxWbRySji4bNpwzjAlMyR07QzqNUMHRpDfrK2cxM3W8r9k6O73kshLDMKBC\nSi61B100uR9ufx2dP9J+XrMQ7mfmkuUoc3yjtXXJtxrWdxjrOTNbDSOiMkquCuQGDc55qyHY7BFc\nmygaxo2KzHMpwenTp/Xw8HBu9zOqSjTH2aR35E0ZysjDWBZE5JKqnnZds5HcDc9455KVeX0uip3g\njBA1xzwWQnZ7zX06piAT00mJOBxZ10qhkBfR4q5WHFfg4PHOxWUE/zXKxB3oebQz9rJuM+2IBVvo\nWgGwuKtGJs6oUjW2dts5Ie+MMghGMlM4izyj0NFSou7+Ujqe0rFji12bFVNyxnTh4eJRx0a2bBDt\nKjA50oGRA5Jmli9dkHbCDItMmImIZx6l0G85zLjYvYrk4czHkTa3Lo7J3C8Y1DtN0TZwmMQuMzJZ\nR40OnZbf5Ax273bEs825NjO+Id5xHGauLieRmZOOtRMG4QmjMA3a9dHnINhQU3va02YYySp+3ZX/\n2BQJza2MtPH7T5gwCZnS341H1QpN7mTajDxUx4GUxvURme5j8zCvLvLukZa7eMCldDlHgnvN1aJt\nkErnagPNqqOcCGvO+si6VgzKjNY1y2FKbhmJHs50p/EqP40e8rrW6zmd0znXklYa8fuN0wbpmpOJ\nsqOfJTtKr5lOmxFC0Ffm6euieMedTsm5626k4Fz3K9gG6fofp4uX253OpfgmCukPYTlleEsXMys5\n4B3gTeD1KDPgDuD7wCD8e3tePqbklhH3JHZ2WMjoOxmxSDPydj/UBZVC2GHTYo1HErl9xZdH7ijR\nNTHuqYspOu50Ss4Xx9f3/eJtUEwpe/LTsbJ3PRP++su+VpSylNydiXMvAM+En58BvpyXjym5JSQV\nXHq2YMquvOMP7zg2bt4ozpdlhmwTAaH9SsOXx5E6okcxetN769txFBqV+Ufi7vT+NihS//468oww\nR2L4Zcxs04JkKblZHA9ngd3w8y5wboa8jAURebXq7UHi4bjo35+u4MR3fz94te5ge200Ob22De2B\nkoysFkw853nXosA/J3EmC2PcBq+3Ba/QpSf+/XlkyeC95nHaFCvP9KS9kDXObNaBA/YupKf7p2mD\nIvJ6yxXFK3GuO8xqt5w2LYGiSk6B74nIJRGJIoXereO4Dj8D7i5dOuOYGXJhL/BqbZ4p/oRFHadY\nB66HHaqA8szzrkUdaWKpS5r1HUUHwfKX7vMJT19uHg4Zhh2eD8KgkYw35K6LPvue9KzvpEYawVZZ\nYczg+JHazGHcXnEZa2c2c5b6FG2Dot7NdLronWrnD19WnRds01koquQeU9VTwDrwBRF5PH4xHC46\n3w+bV3Bp4ygMCH6Yp/kVjX55HR24MEM6nQJrK4Z9+vFe61vqMuzQSC5bqJ1hs+7Ic5rlMiFRBx7v\n+jK6cVgXiU7vXHdYBp72qj3ISYCDKwwK53XENnCm6bB/Oahs5w9fVp0foT2mpZCSU9Xr4d8PgO8Q\nBJJ+P4q9Gv79wPPduQSXNo5Afz/c426azhh2tALfWT/XJGVGDTs0ZI29vWsT686itKMwi/0WsrbB\nxou+jhisR2v1CTrKwTYTScPdX7I7TywPog0OJkM9RgGH6u2BI65IVBebOAfCYSzdfquk16287bXO\nudBET27MME0buMJcJttg7UTCNO63kPNwbhNGyn7YoeFdgxiswXOvH8y6NgO+ybroAG4Fbot9PgCe\nAF5k0vHwQl5e5nhYIlIT4AW9e1O6+8eT3BmT6a60mV7dyYnq1D3I8rK680iVb0YHzMg7XcDbW8i7\nmmwvj6PDVXdF2yCVzr0WZLweLp4mXm+p/ON1nixn1rXikOF4yN2FREQeIBi9QRD45u9V9a9E5OPA\nHvA7wE+ATVX9r6y8bBcSw0gz7DRY2z5pu6HMQNYuJLbVkmEYK49ttWQYxg2LKTnDMCqNKTnDMCqN\nKTnDMCqNKTnDMCqNKTnDMCqNKTnDMCqNKTnDMCqNKTnDMCqNKTnDMCqNKTnDMCqNKTnDMCqNKTnD\nMCqNKTnDMCrNXLdaEpFfAlfndsPj407g54sWYkaqUAaoRjmqUAZYbDl+V1WdW4/fPGdBrvr2fFol\nRORw1ctRhTJANcpRhTLA8pbDzFXDMCqNKTnDMCrNvJVcd873Oy6qUI4qlAGqUY4qlAGWtBxzdTwY\nhmHMGzNXDcOoNHNTciLyhIhcFZGhiDwzr/vOioi8IyJvisjrInIYnrtDRL4vIoPw7+2LljOJiLwk\nIh+IyOXYOafcEvDXYdu8ISKnFif5GE8ZnhOR62F7vC4iG7FrfxmW4aqI/PFipE4jIveJyA9E5IqI\nvCUiW+H5lWmPjDIsf3v4ArKWeQA3AT8CHgBuAX4InJjHvUuQ/R3gzsS5F5gMrP3lRcvpkPtx4BRw\nOU9uYAPoAwJ8Bnh10fJnlOE54IuOtCfC5+qjwP3h83bTossQynYPcCr8fBtwLZR3ZdojowxL3x7z\nGsk9CgxV9ceq+ivgZeDsnO59HJwFdsPPu8C5BcriRFX/BUgG+/bJfRb4Ww34N+C3ReSe+Ujqx1MG\nH2eBl1X1f1T1P4EhwXO3cFT1PVX9j/DzL4G3gXtZofbIKIOPpWmPeSm5e4Gfxv5/l+wKWiYU+J6I\nXBKRZnjublV9L/z8M+DuxYg2NT65V619ngrNuJdiUwUrUQYR+STwaeBVVrQ9EmWAJW8Pczzk85iq\nngLWgS+IyOPxixqMzVfORb2qcgNfAz4FPAy8B3xlseIUR0Q+BnwL2FbVX8SvrUp7OMqw9O0xLyV3\nHbgv9v8nwnNLj6peD/9+AHyHYMj9fmQ+hH8/WJyEU+GTe2XaR1XfV9Vfq+pvgK8zNoGWugwi8hEC\n5fBNVf12eHql2sNVhlVoj3kpudeANRG5X0RuAZ4EXpnTvY+MiNwqIrdFn4E/Ai4TyH4+THYe+O5i\nJJwan9yvAH8WevU+A/x3zIxaKhJzU58jaA8IyvCkiHxURO4H1oB/n7d8LkREgG8Ab6vqV2OXVqY9\nfGVYifaYo3dmg8Aj8yPgS4vwshxB5gcIPEQ/BN6K5AY+DvwzMAD+Cbhj0bI6ZP8HAvPhfwnmQ/7c\nJzeBF+9vwrZ5Ezi9aPkzyvB3oYxvEHSke2LpvxSW4Sqwvmj5Y3I9RmCKvgG8Hh4bq9QeGWVY+vaw\nNx4Mw6g05ngwDKPSmJIzDKPSmJIzDKPSmJIzDKPSmJIzDKPSmJIzDKPSmJIzDKPSmJIzDKPS/D+c\nchS8zJDIoAAAAABJRU5ErkJggg==\n","text/plain":["<Figure size 360x216 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"JFscOzQvqvJZ","colab_type":"code","colab":{}},"source":["from sklearn.metrics import average_precision_score"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"8ZoxBVbN8a65","colab_type":"code","colab":{}},"source":["# Calculate average precision and the PR curve\n","average_precision = average_precision_score(y_test, predicted)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"wGBi9SPIq_Ol","colab_type":"code","colab":{}},"source":["from sklearn.metrics import precision_recall_curve"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"1d0ss9-KrK8A","colab_type":"code","colab":{}},"source":["# pre-loaded\n","def plot_pr_curve(recall, precision, average_precision):\n","  plt.step(recall, precision, color='b', alpha=0.2, where='post')\n","  plt.fill_between(recall, precision, step='post', alpha=0.2, color='b')\n","  plt.xlabel('Recall')\n","  plt.ylabel('Precision')\n","  plt.ylim([0.0, 1.05])\n","  plt.xlim([0.0, 1.0])\n","  plt.title('2-class Precision-Recall curve: AP={0:0.2f}'.format(average_precision))\n","  plt.show()"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"pyQ_5njF8a_N","colab_type":"code","outputId":"ec19192e-a21c-4f5d-ae6f-67859e8ca3fc","executionInfo":{"status":"ok","timestamp":1573277989979,"user_tz":-480,"elapsed":948,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":295}},"source":["# Obtain precision and recall\n","precision, recall, _ = precision_recall_curve(y_test, predicted)\n","\n","# Plot the recall precision tradeoff\n","plot_pr_curve(recall, precision, average_precision)"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"DtnQ7WYMrTtt","colab_type":"text"},"source":["#### What's the benefit of the performance metric ROC curve (AUROC) versus Precision and Recall?\n","\n","Correct answer:\n","\n","The **AUROC** answers the question: \"**How well can this classifier be expected to perform in general, at a variety of different baseline probabilities?**\" but precision and recall don't.\n","\n","\n","The ROC curve plots the true positives vs. false positives , for a classifier, as its discrimination threshold is varied. Since, a random method describes a horizontal curve through the unit interval, it has an AUC of 0.5. Minimally, classifiers should perform better than this, and the extent to which they score higher than one another (meaning the area under the ROC curve is larger), they have better expected performance."]},{"cell_type":"markdown","metadata":{"id":"iMT1tG6OrknV","colab_type":"text"},"source":["###Adjusting your algorithm weights\n"]},{"cell_type":"code","metadata":{"id":"dj2D78l-8odp","colab_type":"code","outputId":"3bf7f79d-f559-4d85-dbfd-f7648b342932","executionInfo":{"status":"ok","timestamp":1573278089347,"user_tz":-480,"elapsed":746,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":275}},"source":["# Define the model with balanced subsample\n","model = RandomForestClassifier(class_weight='balanced_subsample', random_state=5)\n","\n","# Fit your training model to your training set\n","model.fit(X_train, y_train)\n","\n","# Obtain the predicted values and probabilities from the model \n","predicted = model.predict(X_test)\n","probs = model.predict_proba(X_test)\n","\n","# Print the roc_auc_score, the classification report and confusion matrix\n","print(roc_auc_score(y_test, probs[:,1]))\n","print(classification_report(y_test, predicted))\n","print(confusion_matrix(y_test, predicted))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0.999468438538206\n","              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.90      0.90      0.90        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.95      0.95      0.95      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","[[1504    1]\n"," [   1    9]]\n"],"name":"stdout"},{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/ensemble/forest.py:245: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n","  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"NABtSWFaruVA","colab_type":"text"},"source":["- In the next exercises you'll see how to more smartly tweak your model to focus on reducing false negatives and catch more fraud."]},{"cell_type":"markdown","metadata":{"id":"RfWRVbikrwPT","colab_type":"text"},"source":["####Adjusting your Random Forest to fraud detection\n","\n","You'll define weights manually, to be able to off-set that imbalance slightly. "]},{"cell_type":"code","metadata":{"id":"R9owSmYFr3wD","colab_type":"code","colab":{}},"source":["# pre-defined\n","\n","def get_model_results(X_train, y_train, X_test, y_test, model):\n","  model.fit(X_train, y_train)\n","  predicted = model.predict(X_test)\n","  probs = model.predict_proba(X_test)\n","  print (classification_report(y_test, predicted))\n","  print (confusion_matrix(y_test, predicted))"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"jw_Y0Nmfr9Fq","colab_type":"text"},"source":["- Change the **weight option** to set the ratio to 1 to 12 for the non-fraud and fraud cases, and set the split criterion to 'entropy'.\n","- Set the maximum depth to 10.\n","- Set the minimal samples in leaf nodes to 10.\n","- Set the number of trees to use in the model to 20."]},{"cell_type":"code","metadata":{"id":"eH55OviT8sQN","colab_type":"code","outputId":"2a792799-8390-4f86-8ad0-4cda7fa5f3ec","executionInfo":{"status":"ok","timestamp":1573278194059,"user_tz":-480,"elapsed":1001,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":204}},"source":["# Change the model options\n","model = RandomForestClassifier(bootstrap=True, class_weight={0:1, 1:12}, criterion='entropy',\n","                               \n","\t\t\t# Change depth of model\n","            max_depth=10, \n","                               \n","\t\t\t# Change the number of samples in leaf nodes\n","            min_samples_leaf=10, \n","                               \n","\t\t\t# Change the number of trees to use\n","            n_estimators=20, n_jobs=-1, random_state=5)\n","\n","# Run the function get_model_results\n","get_model_results(X_train, y_train, X_test, y_test, model)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.83      1.00      0.91        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.92      1.00      0.95      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","[[1503    2]\n"," [   0   10]]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"qam0xsZ1sGEE","colab_type":"text"},"source":["**Observation:**\n","\n","You can see by smartly defining more options in the model, you can obtain better predictions. You have effectively reduced the number of false negatives, i.e. you are catching more cases of fraud, whilst keeping the number of false positives low.\n","\n","There is a smarter way of doing it, by using `GridSearchCV`."]},{"cell_type":"markdown","metadata":{"id":"zyS3vXjVsNXh","colab_type":"text"},"source":["####GridSearchCV to find optimal parameters"]},{"cell_type":"code","metadata":{"id":"LU7zaZs-sTRU","colab_type":"code","colab":{}},"source":["from sklearn.model_selection import GridSearchCV"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"By1Av_WV8xIs","colab_type":"code","outputId":"5dc0ef76-5de3-45d0-cd2e-b01728dd12c2","executionInfo":{"status":"ok","timestamp":1573278324093,"user_tz":-480,"elapsed":9073,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":85}},"source":["# Define the parameter sets to test\n","param_grid = {'n_estimators': [1, 30], 'max_features': ['auto', 'log2'], 'max_depth': [4, 8], 'criterion': ['gini', 'entropy']\n","}\n","\n","# Define the model to use\n","model = RandomForestClassifier(random_state=5)\n","\n","# Combine the parameter sets with the defined model\n","CV_model = GridSearchCV(estimator=model, param_grid=param_grid, cv=5, scoring='recall', n_jobs=-1)\n","\n","# Fit the model to our training data and obtain best parameters\n","CV_model.fit(X_train, y_train)\n","CV_model.best_params_"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["{'criterion': 'gini',\n"," 'max_depth': 4,\n"," 'max_features': 'log2',\n"," 'n_estimators': 30}"]},"metadata":{"tags":[]},"execution_count":65}]},{"cell_type":"markdown","metadata":{"id":"LULMqQUNs6x6","colab_type":"text"},"source":["####Model results using GridSearchCV\n"]},{"cell_type":"code","metadata":{"id":"wuHuGA8E9DX5","colab_type":"code","colab":{}},"source":["def get_model_results(X_train, y_train, X_test, y_test, model):\n","  model.fit(X_train, y_train)\n","  predicted = model.predict(X_test)\n","  probs = model.predict_proba(X_test)\n","  print (classification_report(y_test, predicted))\n","  print (confusion_matrix(y_test, predicted))"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"GWC0-FZu85Sz","colab_type":"code","outputId":"0d39e9aa-89ba-46d8-c04c-271402ad5d91","executionInfo":{"status":"ok","timestamp":1573278343755,"user_tz":-480,"elapsed":1100,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":204}},"source":["# Input the optimal parameters in the model\n","model = RandomForestClassifier(class_weight={0:1,1:12}, criterion='gini',\n","            max_depth=8, max_features='log2', min_samples_leaf=10, n_estimators=30, n_jobs=-1, random_state=5)\n","\n","# Get results from your model\n","get_model_results(X_train, y_train, X_test, y_test, model)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.90      0.90      0.90        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.95      0.95      0.95      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","[[1504    1]\n"," [   1    9]]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"azmKT9OrLgB4","colab_type":"text"},"source":["### Ensemble methods\n"]},{"cell_type":"markdown","metadata":{"id":"DUOTaIqvtIgz","colab_type":"text"},"source":["####Logistic Regression\n"]},{"cell_type":"code","metadata":{"id":"eZ58Ax718xLe","colab_type":"code","outputId":"af540369-4177-4c24-841a-cdb4c8fc8466","executionInfo":{"status":"ok","timestamp":1573278539190,"user_tz":-480,"elapsed":720,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":258}},"source":["# Define the Logistic Regression model with weights\n","model = LogisticRegression(class_weight={0:1, 1:15}, random_state=5)\n","\n","# Get the model results\n","get_model_results(X_train, y_train, X_test, y_test, model)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.57      0.80      0.67        10\n","\n","    accuracy                           0.99      1515\n","   macro avg       0.79      0.90      0.83      1515\n","weighted avg       1.00      0.99      1.00      1515\n","\n","[[1499    6]\n"," [   2    8]]\n"],"name":"stdout"},{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n","  FutureWarning)\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"bpPhleBKtZ0s","colab_type":"text"},"source":["####Voting Classifier\n"]},{"cell_type":"code","metadata":{"id":"kdUzGz9Mtlv-","colab_type":"code","colab":{}},"source":["from sklearn.tree import DecisionTreeClassifier"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"1v0qs3xHtuAK","colab_type":"code","colab":{}},"source":["def get_model_results(X_train, y_train, X_test, y_test, model):\n","  model.fit(X_train, y_train)\n","  predicted = model.predict(X_test) # for classification\n","  #probs = model.predict_proba(X_test)\n","  print (classification_report(y_test, predicted))\n","  print (confusion_matrix(y_test, predicted))"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"HggllXF3Lm0h","colab_type":"code","outputId":"75b9658b-0503-4d7a-cba6-cdd171d84d4b","executionInfo":{"status":"ok","timestamp":1573278645162,"user_tz":-480,"elapsed":2650,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":258}},"source":["# Import the package\n","from sklearn.ensemble import VotingClassifier\n","\n","# Define the three classifiers to use in the ensemble\n","clf1 = LogisticRegression(class_weight={0:1, 1:15}, random_state=5)\n","clf2 = RandomForestClassifier(class_weight={0:1, 1:12}, criterion='gini', max_depth=8, max_features='log2',\n","                              min_samples_leaf=10, n_estimators=30, n_jobs=-1, random_state=5)\n","clf3 = DecisionTreeClassifier(random_state=5, class_weight=\"balanced\")\n","\n","# Combine the classifiers in the ensemble model\n","ensemble_model = VotingClassifier(estimators=[('lr', clf1), ('rf', clf2), ('dt', clf3)], voting='hard')\n","\n","# Get the results \n","get_model_results(X_train, y_train, X_test, y_test, ensemble_model)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n","  FutureWarning)\n"],"name":"stderr"},{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.89      0.80      0.84        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.94      0.90      0.92      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","[[1504    1]\n"," [   2    8]]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"nQucpQOLt6ZF","colab_type":"text"},"source":["####Adjust weights within the Voting Classifier"]},{"cell_type":"code","metadata":{"id":"S3kDH-g4M25z","colab_type":"code","outputId":"e26ecb51-3baa-48cd-f8f1-83f68a52935e","executionInfo":{"status":"ok","timestamp":1573278709817,"user_tz":-480,"elapsed":2678,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":258}},"source":["# Define the ensemble model\n","ensemble_model = VotingClassifier(estimators=[('lr', clf1), ('rf', clf2), ('gnb', clf3)], voting='soft', weights=[1, 4, 1], flatten_transform=True)\n","\n","# Get results \n","get_model_results(X_train, y_train, X_test, y_test, ensemble_model)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n","  FutureWarning)\n"],"name":"stderr"},{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","         0.0       1.00      1.00      1.00      1505\n","         1.0       0.91      1.00      0.95        10\n","\n","    accuracy                           1.00      1515\n","   macro avg       0.95      1.00      0.98      1515\n","weighted avg       1.00      1.00      1.00      1515\n","\n","[[1504    1]\n"," [   0   10]]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"eh5Vl8DyumX_","colab_type":"text"},"source":["## 3. Fraud detection using unlabelled data\n","\n","This chapter focuses on using unsupervised learning techniques to detect fraud. You will segment customers, use K-means clustering and other clustering algorithms to find suspicious occurrences in your data."]},{"cell_type":"markdown","metadata":{"id":"2NRQbpztuEPX","colab_type":"text"},"source":["###Normal versus abnormal behavior\n"]},{"cell_type":"code","metadata":{"id":"05Ay20fhuxlh","colab_type":"code","colab":{}},"source":["df = pd.read_csv(\"banksim.csv\")\n","del df['Unnamed: 0']"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"ixVrpGnG8sTp","colab_type":"code","outputId":"fe2ea49b-a3a5-47c6-a1f6-f2748aa396ba","executionInfo":{"status":"ok","timestamp":1573280807145,"user_tz":-480,"elapsed":885,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# Get the dataframe shape\n","df.shape"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(7200, 5)"]},"metadata":{"tags":[]},"execution_count":92}]},{"cell_type":"code","metadata":{"id":"TsQ9N3s0u6X5","colab_type":"code","outputId":"95ad0ea8-a825-44db-e4ed-9ddd83c8449f","executionInfo":{"status":"ok","timestamp":1573280808405,"user_tz":-480,"elapsed":1150,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":204}},"source":["# Display the first 5 rows\n","df.head()"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>age</th>\n","      <th>gender</th>\n","      <th>category</th>\n","      <th>amount</th>\n","      <th>fraud</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>3</td>\n","      <td>F</td>\n","      <td>es_transportation</td>\n","      <td>49.71</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>4</td>\n","      <td>F</td>\n","      <td>es_health</td>\n","      <td>39.29</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>3</td>\n","      <td>F</td>\n","      <td>es_transportation</td>\n","      <td>18.76</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>4</td>\n","      <td>M</td>\n","      <td>es_transportation</td>\n","      <td>13.95</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>2</td>\n","      <td>M</td>\n","      <td>es_transportation</td>\n","      <td>49.87</td>\n","      <td>0</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["  age gender           category  amount  fraud\n","0   3      F  es_transportation   49.71      0\n","1   4      F          es_health   39.29      0\n","2   3      F  es_transportation   18.76      0\n","3   4      M  es_transportation   13.95      0\n","4   2      M  es_transportation   49.87      0"]},"metadata":{"tags":[]},"execution_count":93}]},{"cell_type":"code","metadata":{"id":"4RfWp0fB8htp","colab_type":"code","outputId":"4bfb3417-e535-481b-cd18-f913cd491f90","executionInfo":{"status":"ok","timestamp":1573280809286,"user_tz":-480,"elapsed":650,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":306}},"source":["# Groupby categories and take the mean \n","print(df.groupby('category').mean())"],"execution_count":0,"outputs":[{"output_type":"stream","text":["                           amount     fraud\n","category                                   \n","es_barsandrestaurants   43.841793  0.022472\n","es_contents             55.170000  0.000000\n","es_fashion              59.780769  0.020619\n","es_food                 35.216050  0.000000\n","es_health              126.604704  0.242798\n","es_home                120.688317  0.208333\n","es_hotelservices       172.756245  0.548387\n","es_hyper                46.788180  0.125000\n","es_leisure             229.757600  1.000000\n","es_otherservices       149.648960  0.600000\n","es_sportsandtoys       157.251737  0.657895\n","es_tech                132.852862  0.179487\n","es_transportation       27.422014  0.000000\n","es_travel              231.818656  0.944444\n","es_wellnessandbeauty    66.167078  0.060606\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"Ek4jz37HvL0W","colab_type":"text"},"source":["**Observation:**\n","\n","The majority of fraud is observed in travel, leisure and sports related transactions."]},{"cell_type":"markdown","metadata":{"id":"y6yewo0PVdp8","colab_type":"text"},"source":["####Customer segmentation\n"]},{"cell_type":"code","metadata":{"id":"-wi64oMk8hwW","colab_type":"code","outputId":"6d61184e-5d3e-48c5-efeb-5f9926d22562","executionInfo":{"status":"ok","timestamp":1573280811337,"user_tz":-480,"elapsed":829,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":187}},"source":["# Group by age groups and get the mean\n","print(df.groupby('age').mean())"],"execution_count":0,"outputs":[{"output_type":"stream","text":["        amount     fraud\n","age                     \n","0    49.468935  0.050000\n","1    35.622829  0.026648\n","2    37.228665  0.028718\n","3    37.279338  0.023283\n","4    36.197985  0.035966\n","5    37.547521  0.023990\n","6    36.700852  0.022293\n","U    39.117000  0.000000\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"63Rl1ufNNJky","colab_type":"code","outputId":"90ae1e4c-ca97-4972-c7fa-4acf7045067b","executionInfo":{"status":"ok","timestamp":1573280812914,"user_tz":-480,"elapsed":711,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":170}},"source":["# Count the values of the observations in each age group\n","print(df['age'].value_counts())"],"execution_count":0,"outputs":[{"output_type":"stream","text":["2    2333\n","3    1718\n","4    1279\n","5     792\n","1     713\n","6     314\n","0      40\n","U      11\n","Name: age, dtype: int64\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"GA5LbXBhvecP","colab_type":"text"},"source":["**Observation:**\n","\n","As you can see the average amount spent as well as fraud occurrence is rather **similar across groups**. Age group '0' stands out but since there are only 40 cases, it does not make sense to split these out in a separate group and run a separate model on them."]},{"cell_type":"markdown","metadata":{"id":"AagwWi-X2b4g","colab_type":"text"},"source":["####Using statistics to define normal behavior"]},{"cell_type":"code","metadata":{"id":"hHsPBXQYNStm","colab_type":"code","outputId":"5694a13d-8a9f-4dd3-e6a8-71be2d4bad88","executionInfo":{"status":"ok","timestamp":1573281001785,"user_tz":-480,"elapsed":768,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":265}},"source":["# Create two dataframes with fraud and non-fraud data \n","df_fraud = df.loc[df.fraud == 1] \n","df_non_fraud = df.loc[df.fraud == 0]\n","\n","# Plot histograms of the amounts in fraud and non-fraud data \n","plt.hist(df_fraud.amount, alpha=0.5, label='fraud')\n","plt.hist(df_non_fraud.amount, alpha=0.5, label='nonfraud')\n","plt.legend()\n","plt.show()"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAX0AAAD4CAYAAAAAczaOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"Kxw9GjceNU10","colab_type":"text"},"source":["### Clustering methods to detect fraud\n"]},{"cell_type":"code","metadata":{"id":"wskHf44d3BFx","colab_type":"code","colab":{}},"source":["# pre-loaded\n","df = pd.read_csv(\"banksim_adj.csv\")\n","\n","y = df.fraud.values.astype(np.float)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"GR5FGKXU4ymy","colab_type":"code","outputId":"0a5ac560-fa2b-4818-ed24-386efa519b34","executionInfo":{"status":"ok","timestamp":1573281572521,"user_tz":-480,"elapsed":701,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# match dataset\n","df = df.drop(['Unnamed: 0', 'fraud'], axis=1)\n","df.shape"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(7189, 17)"]},"metadata":{"tags":[]},"execution_count":111}]},{"cell_type":"markdown","metadata":{"id":"Y7VOLily33LU","colab_type":"text"},"source":["####Scaling the data\n"]},{"cell_type":"code","metadata":{"id":"WgSDQgMmNYmg","colab_type":"code","colab":{}},"source":["# Import the scaler\n","from sklearn.preprocessing import MinMaxScaler\n","\n","# Take the float values of df for X\n","X = df.values.astype(np.float)\n","\n","# Define the scaler and apply to the data\n","scaler = MinMaxScaler()\n","X_scaled = scaler.fit_transform(X)"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"d3PtSFWP354K","colab_type":"text"},"source":["####K-means clustering\n"]},{"cell_type":"code","metadata":{"id":"5hHP_pVMNSxU","colab_type":"code","outputId":"6ee25fed-7685-4c40-996e-1d2fb78a25af","executionInfo":{"status":"ok","timestamp":1573281313296,"user_tz":-480,"elapsed":733,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":85}},"source":["# Import MiniBatchKmeans \n","from sklearn.cluster import MiniBatchKMeans\n","\n","# Define the model \n","kmeans = MiniBatchKMeans(n_clusters=8, random_state=0)\n","\n","# Fit the model to the scaled data\n","kmeans.fit(X_scaled)"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["MiniBatchKMeans(batch_size=100, compute_labels=True, init='k-means++',\n","                init_size=None, max_iter=100, max_no_improvement=10,\n","                n_clusters=8, n_init=3, random_state=0, reassignment_ratio=0.01,\n","                tol=0.0, verbose=0)"]},"metadata":{"tags":[]},"execution_count":103}]},{"cell_type":"markdown","metadata":{"id":"W4jpI7tj4B7s","colab_type":"text"},"source":["####Elbow method\n","\n","To decide which amount of clusters you're going to use, let's apply the Elbow method and see what the optimal number of clusters should be based on this method."]},{"cell_type":"code","metadata":{"id":"jSUlA4wrNb6L","colab_type":"code","outputId":"00e236bf-0428-4a39-fb46-cac8b27915bc","executionInfo":{"status":"ok","timestamp":1573281367207,"user_tz":-480,"elapsed":1891,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":295}},"source":["# Define the range of clusters to try\n","clustno = range(1, 10)\n","\n","# Run MiniBatch Kmeans over the number of clusters\n","kmeans = [MiniBatchKMeans(n_clusters=i) for i in clustno]\n","\n","# Obtain the score for each model\n","score = [kmeans[i].fit(X_scaled).score(X_scaled) for i in range(len(kmeans))]\n","\n","# Plot the models and their respective score \n","plt.plot(clustno, score)\n","plt.xlabel('Number of Clusters')\n","plt.ylabel('Score')\n","plt.title('Elbow Curve')\n","plt.show()"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAZMAAAEWCAYAAACjYXoKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nNZvDxMzMms1hYmZmzfb/AbkZVlNXG/hqAAAAAElFTkSuQmCC\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"15Sgdtz9Ngfv","colab_type":"text"},"source":["### Assigning fraud versus non-fraud\n"]},{"cell_type":"markdown","metadata":{"id":"_StvaiMT4OEG","colab_type":"text"},"source":["####Detecting outliers\n","\n","The fraudulent transactions are typically flagged as the observations that are furthest aways from the cluster centroid. \n","\n","You'll learn how to do this and how to determine the cut-off in this exercise. "]},{"cell_type":"code","metadata":{"id":"GEizNutV4aql","colab_type":"code","outputId":"afecef51-3f83-4da9-dc15-16e66779f4e0","executionInfo":{"status":"ok","timestamp":1573281493729,"user_tz":-480,"elapsed":754,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["len(X_scaled) # 7189"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["7189"]},"metadata":{"tags":[]},"execution_count":108}]},{"cell_type":"code","metadata":{"id":"LKb2UR9b4YQq","colab_type":"code","outputId":"671d8005-b37c-4102-fd84-4d061ff33c04","executionInfo":{"status":"ok","timestamp":1573281588100,"user_tz":-480,"elapsed":713,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["len(y) # 7189"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["7189"]},"metadata":{"tags":[]},"execution_count":114}]},{"cell_type":"code","metadata":{"id":"94JGUTSWNe71","colab_type":"code","colab":{}},"source":["# Split the data into training and test set\n","X_train, X_test, y_train, y_test = train_test_split(X_scaled, y, test_size=0.3, random_state=0)\n","\n","# Define K-means model \n","kmeans = MiniBatchKMeans(n_clusters=3, random_state=42).fit(X_train)\n","\n","# Obtain predictions and calculate distance from cluster centroid\n","X_test_clusters = kmeans.predict(X_test)\n","X_test_clusters_centers = kmeans.cluster_centers_\n","dist = [np.linalg.norm(x-y) for x, y in zip(X_test, X_test_clusters_centers[X_test_clusters])]\n","\n","# Create fraud predictions based on outliers on clusters \n","km_y_pred = np.array(dist)\n","km_y_pred[dist >= np.percentile(dist, 95)] = 1\n","km_y_pred[dist < np.percentile(dist, 95)] = 0"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"Lcxmyqwp5Jo6","colab_type":"text"},"source":["####Checking model results\n"]},{"cell_type":"code","metadata":{"id":"4ExCesS2NpSH","colab_type":"code","outputId":"5b70611b-d7e6-4f9f-f359-46682c2960a4","executionInfo":{"status":"ok","timestamp":1573281594024,"user_tz":-480,"elapsed":726,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# Obtain the ROC score\n","print(roc_auc_score(y_test, km_y_pred))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0.8109115999408585\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"dFFXNI1i5U2D","colab_type":"code","colab":{}},"source":["# pre-defined\n","def plot_confusion_matrix(km_cm):\n","  df_cm = pd.DataFrame(km_cm, ['True Normal','True Fraud'],['Pred Normal','Pred Fraud'])\n","  plt.figure(figsize = (8,4))\n","  sn.set(font_scale=1.4) \n","  sn.heatmap(df_cm, annot=True,annot_kws={\"size\": 16},fmt='g')\n","  plt.show()"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"Ch3RxICR5bYI","colab_type":"code","colab":{}},"source":["import seaborn as sn"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"WkhO30OPNe5M","colab_type":"code","outputId":"f75215bd-b70b-4511-d182-d2eb46a6f727","executionInfo":{"status":"ok","timestamp":1573281716059,"user_tz":-480,"elapsed":930,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":334}},"source":["# Create a confusion matrix\n","km_cm = confusion_matrix(y_test, km_y_pred)\n","\n","# Plot the confusion matrix in a figure to visualize results \n","plot_confusion_matrix(km_cm)"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAd0AAAE9CAYAAABOebjsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n","text/plain":["<Figure size 576x288 with 2 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"dYH3dWeI5mT3","colab_type":"text"},"source":["**Observation:**\n","\n","By lowering the threshold for cases flagged as fraud, you flag more cases overall but thereby also get more false positives."]},{"cell_type":"markdown","metadata":{"id":"sU5Rf7ZBNv-0","colab_type":"text"},"source":["### Other clustering fraud detection methods\n"]},{"cell_type":"markdown","metadata":{"id":"8eIfn76K6lgu","colab_type":"text"},"source":["#### DB Scan\n","\n","The advantage of DBSCAN is that you do **not need to define the number of clusters** beforehand. Also, DBSCAN can handle **weirdly shaped** data (i.e. non-convex) much better than K-means can. \n","\n","This time, you are not going to take the outliers of the clusters and use that for fraud, but take the smallest clusters in the data and label those as fraud."]},{"cell_type":"code","metadata":{"id":"jmsBPbJ97J3x","colab_type":"code","colab":{}},"source":["import pickle\n","file = open(\"labels_full.pickle\", \"rb\")\n","labels = pickle.load(file)\n","file.close()"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"mbJ2Hn4L8Dc7","colab_type":"code","outputId":"52175391-74ad-46ee-ea0e-dad9c2745b4a","executionInfo":{"status":"ok","timestamp":1573282541963,"user_tz":-480,"elapsed":467,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":221}},"source":["# datacamp used a subsample of this\n","labels"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["0       0\n","1       0\n","2       0\n","3       0\n","4       0\n","       ..\n","7184    1\n","7185    1\n","7186    1\n","7187    1\n","7188    1\n","Name: fraud, Length: 7189, dtype: int64"]},"metadata":{"tags":[]},"execution_count":136}]},{"cell_type":"code","metadata":{"id":"a-D7I5Mq8OgP","colab_type":"code","colab":{}},"source":["from sklearn.metrics.cluster.supervised import homogeneity_score\n","from sklearn.metrics.cluster.unsupervised import silhouette_score"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"EyjV5-Jb8eou","colab_type":"code","outputId":"e3426bb4-e6c9-4302-fd37-9d9913d3066c","executionInfo":{"status":"ok","timestamp":1573282521926,"user_tz":-480,"elapsed":718,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["len(pred_labels)"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["7189"]},"metadata":{"tags":[]},"execution_count":134}]},{"cell_type":"code","metadata":{"id":"ift6gz2TNw-u","colab_type":"code","outputId":"30a811d0-5be3-4ab1-9936-0995393a5301","executionInfo":{"status":"ok","timestamp":1573282606818,"user_tz":-480,"elapsed":3283,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":68}},"source":["# Import DBSCAN\n","from sklearn.cluster import DBSCAN\n","\n","# Initialize and fit the DBscan model\n","db = DBSCAN(eps=0.9, min_samples=10, n_jobs=-1).fit(X_scaled)\n","\n","# Obtain the predicted labels and calculate number of clusters\n","pred_labels = db.labels_\n","n_clusters = len(set(pred_labels)) - (1 if -1 in labels else 0)\n","\n","# Print performance metrics for DBscan\n","print('Estimated number of clusters: %d' % n_clusters)\n","print(\"Homogeneity: %0.3f\" % homogeneity_score(labels, pred_labels))\n","print(\"Silhouette Coefficient: %0.3f\" % silhouette_score(X_scaled, pred_labels))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Estimated number of clusters: 23\n","Homogeneity: 0.612\n","Silhouette Coefficient: 0.713\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"gpU7GJDa9A0Q","colab_type":"text"},"source":["####Assessing smallest clusters\n","\n","In this exercise you're going to have a look at the clusters that came out of DBscan, and flag certain clusters as fraud:\n","\n","- you first need to figure out how big the clusters are, and filter out the smallest\n","- then, you're going to take the smallest ones and flag those as fraud\n","- last, you'll check with the original labels whether this does actually do a good job in detecting fraud."]},{"cell_type":"code","metadata":{"id":"_oZZREG9Nw8s","colab_type":"code","outputId":"7ab3df74-4edc-4c93-8c3e-8a557c8121be","executionInfo":{"status":"ok","timestamp":1573282629322,"user_tz":-480,"elapsed":783,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":51}},"source":["# Count observations in each cluster number\n","counts = np.bincount(pred_labels[pred_labels >= 0])\n","\n","# Print the result\n","print(counts)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["[3252  145 2714   55  174  119  122   98   54   15   76   15   43   25\n","   51   47   42   15   25   20   19   10]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"3HLE7MHPNtSW","colab_type":"code","outputId":"fdb2cd58-7d37-4a1c-811c-c96d3497797e","executionInfo":{"status":"ok","timestamp":1573282632069,"user_tz":-480,"elapsed":866,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":51}},"source":["# Sort the sample counts of the clusters and take the top 3 smallest clusters\n","smallest_clusters = np.argsort(counts)[:3]\n","\n","# Print the results \n","print(\"The smallest clusters are clusters:\")      \n","print(smallest_clusters)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["The smallest clusters are clusters:\n","[21 17  9]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"wEEdpcYLOG1y","colab_type":"code","outputId":"45c3f532-a72c-41de-8deb-c9f8205e48ad","executionInfo":{"status":"ok","timestamp":1573282633969,"user_tz":-480,"elapsed":975,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":51}},"source":["# Print the counts of the smallest clusters only\n","print(\"Their counts are:\")      \n","print(counts[smallest_clusters])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Their counts are:\n","[10 15 15]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"oJqmc7EM9N0j","colab_type":"text"},"source":["#### Checking results\n"]},{"cell_type":"code","metadata":{"id":"1k2iSk2fNtVh","colab_type":"code","outputId":"92443de3-074e-4124-de30-d5ebbe7c4ead","executionInfo":{"status":"ok","timestamp":1573282635734,"user_tz":-480,"elapsed":747,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":85}},"source":["# Create a dataframe of the predicted cluster numbers and fraud labels\n","df = pd.DataFrame({'clusternr':pred_labels,'fraud':labels})\n","\n","# Create a condition flagging fraud for the smallest clusters \n","df['predicted_fraud'] = np.where((df['clusternr']==21) | (df['clusternr']==17) | (df['clusternr']==9), 1, 0)\n","\n","# Run a crosstab on the results \n","print(pd.crosstab(df['fraud'], df['predicted_fraud'], rownames=['Actual Fraud'], colnames=['Flagged Fraud']))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Flagged Fraud     0   1\n","Actual Fraud           \n","0              6973  16\n","1               176  24\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"eTrU_Tw3mH6J","colab_type":"text"},"source":["## 4. Fraud detection using text\n","\n","\n","Fraud detection using text\n","100%\n"]},{"cell_type":"markdown","metadata":{"colab_type":"text","id":"yUQzlXTH9WYT"},"source":["###Using text data\n"]},{"cell_type":"markdown","metadata":{"id":"O53OSG689iU5","colab_type":"text"},"source":["####Word search with dataframes\n"]},{"cell_type":"markdown","metadata":{"id":"JfPucbj898LC","colab_type":"text"},"source":["In this exercise you're going to work with text data, containing emails from Enron employees. The Enron scandal is a famous fraud case. Enron employees covered up the bad financial position of the company, thereby keeping the stock price artificially high. Enron employees sold their own stock options, and when the truth came out, Enron investors were left with nothing. The goal is to find all emails that mention specific words, such as \"sell enron stock\".\n","\n"]},{"cell_type":"code","metadata":{"id":"ER1xpSYJ9xJo","colab_type":"code","outputId":"73b94c13-3f16-48ab-aaef-7d2ac139bb5d","executionInfo":{"status":"ok","timestamp":1573284295629,"user_tz":-480,"elapsed":709,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["# pre-loaded\n","df = pd.read_csv(\"enron_emails_clean.csv\")\n","df.shape"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(2090, 6)"]},"metadata":{"tags":[]},"execution_count":4}]},{"cell_type":"markdown","metadata":{"id":"1dYTiHkE99qn","colab_type":"text"},"source":["By using string operations on dataframes, you can easily sift through messy email data and create flags based on word-hits. The Enron email data has been put into a dataframe called df so let's search for suspicious terms. "]},{"cell_type":"code","metadata":{"id":"1lPUled2WGTo","colab_type":"code","colab":{}},"source":["# Find all cleaned emails that contain 'sell enron stock'\n","mask = df['clean_content'].str.contains('sell enron stock', na=False)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"AsZvnMJOWow1","colab_type":"code","outputId":"4ac27d8d-8189-4ac9-dea1-a92d940b5630","executionInfo":{"status":"ok","timestamp":1573284301784,"user_tz":-480,"elapsed":715,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":85}},"source":["# Select the data from df that contain the searched for words \n","print(df.loc[mask])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["                                       Message-ID  ...                                      clean_content\n","154  <6336501.1075841154311.JavaMail.evans@thyme>  ...  joint venture enron meeting belies officers cl...\n","\n","[1 rows x 6 columns]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"zDh8NdgG-ETp","colab_type":"text"},"source":["####Using list of terms\n"]},{"cell_type":"code","metadata":{"id":"eKqCqIiLWot_","colab_type":"code","outputId":"2e9604be-9f5e-4624-85e1-67d79488bfa7","executionInfo":{"status":"ok","timestamp":1573284308078,"user_tz":-480,"elapsed":716,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":255}},"source":["# Create a list of terms to search for\n","searchfor = ['enron stock', 'sell stock', 'stock bonus', 'sell enron stock']\n","\n","# Filter cleaned emails on searchfor list and select from df \n","filtered_emails = df.loc[df['clean_content'].str.contains('|'.join(searchfor), na=False)]\n","print(filtered_emails)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["                                         Message-ID  ...                                      clean_content\n","0      <8345058.1075840404046.JavaMail.evans@thyme>  ...  investools advisory free digest trusted invest...\n","1      <1512159.1075863666797.JavaMail.evans@thyme>  ...  forwarded richard b sanders hou ect pm justin ...\n","2     <26118676.1075862176383.JavaMail.evans@thyme>  ...  hey wearing target purple shirt today mine wan...\n","3     <10369289.1075860831062.JavaMail.evans@thyme>  ...  leslie milosevich santa clara avenue alameda c...\n","4     <26728895.1075860815046.JavaMail.evans@thyme>  ...  rini twait e th ave longmont co rtwait graphic...\n","...                                             ...  ...                                                ...\n","1151  <15875618.1075860830584.JavaMail.evans@thyme>  ...  blanca anderson cadiz new orleans la bandersn ...\n","1450  <30798399.1075841348382.JavaMail.evans@thyme>  ...  enron announced today common stock traded coun...\n","1473    <957052.1075861359136.JavaMail.evans@thyme>  ...  today announced plans merge dynegy major playe...\n","1557  <18936682.1075861158419.JavaMail.evans@thyme>  ...  updated merger q document enron updates site h...\n","1621   <5472336.1075841501893.JavaMail.evans@thyme>  ...  please read important information concerning e...\n","\n","[314 rows x 6 columns]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"UmNBWsPa-Pw0","colab_type":"text"},"source":["####Creating a flag\n","\n","This time you are going to create an actual flag variable that gives a `1` when the emails get a hit on the search terms of interest, and `0` otherwise. \n"]},{"cell_type":"code","metadata":{"id":"yOCNKSEgWuiE","colab_type":"code","outputId":"60b940d0-eaea-4309-8691-27de0f4707d2","executionInfo":{"status":"ok","timestamp":1573282960225,"user_tz":-480,"elapsed":803,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":68}},"source":["# Create flag variable where the emails match the searchfor terms\n","df['flag'] = np.where((df['clean_content'].str.contains('|'.join(searchfor)) == True), 1, 0)\n","\n","# Count the values of the flag variable\n","count = df['flag'].value_counts()\n","print(count)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["0    1776\n","1     314\n","Name: flag, dtype: int64\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"DSO6xgtoWyyJ","colab_type":"text"},"source":["###Text mining to detect fraud\n"]},{"cell_type":"markdown","metadata":{"id":"iIjq2pPD-b88","colab_type":"text"},"source":["####Removing stopwords\n","\n","In a first step you need to define the list of stopwords and punctuations that are to be removed in the next exercise from the text data. \n"]},{"cell_type":"code","metadata":{"id":"GPYrlw0CDYMh","colab_type":"code","outputId":"aa8d4149-6626-4336-99d6-c0ba47a714fb","executionInfo":{"status":"ok","timestamp":1573284421406,"user_tz":-480,"elapsed":704,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":289}},"source":["# subsample datasetdf \n","df = df[:5]\n","# match dataset\n","df = df.drop('Message-ID', axis=1)\n","df.head()"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>From</th>\n","      <th>To</th>\n","      <th>Date</th>\n","      <th>content</th>\n","      <th>clean_content</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>('advdfeedback@investools.com')</td>\n","      <td>('advdfeedback@investools.com')</td>\n","      <td>2002-01-29 23:20:55</td>\n","      <td>INVESTools Advisory\\nA Free Digest of Trusted ...</td>\n","      <td>investools advisory free digest trusted invest...</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>('richard.sanders@enron.com')</td>\n","      <td>('richard.sanders@enron.com')</td>\n","      <td>2000-09-20 19:07:00</td>\n","      <td>----- Forwarded by Richard B Sanders/HOU/ECT o...</td>\n","      <td>forwarded richard b sanders hou ect pm justin ...</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>('m..love@enron.com')</td>\n","      <td>('m..love@enron.com')</td>\n","      <td>2001-10-30 16:15:17</td>\n","      <td>hey you are not wearing your target purple shi...</td>\n","      <td>hey wearing target purple shirt today mine wan...</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>('leslie.milosevich@kp.org')</td>\n","      <td>('leslie.milosevich@kp.org')</td>\n","      <td>2002-01-30 17:54:18</td>\n","      <td>Leslie Milosevich\\n1042 Santa Clara Avenue\\nAl...</td>\n","      <td>leslie milosevich santa clara avenue alameda c...</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>('rtwait@graphicaljazz.com')</td>\n","      <td>('rtwait@graphicaljazz.com')</td>\n","      <td>2002-01-30 19:36:01</td>\n","      <td>Rini Twait\\n1010 E 5th Ave\\nLongmont, CO 80501...</td>\n","      <td>rini twait e th ave longmont co rtwait graphic...</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["                              From  ...                                      clean_content\n","0  ('advdfeedback@investools.com')  ...  investools advisory free digest trusted invest...\n","1    ('richard.sanders@enron.com')  ...  forwarded richard b sanders hou ect pm justin ...\n","2            ('m..love@enron.com')  ...  hey wearing target purple shirt today mine wan...\n","3     ('leslie.milosevich@kp.org')  ...  leslie milosevich santa clara avenue alameda c...\n","4     ('rtwait@graphicaljazz.com')  ...  rini twait e th ave longmont co rtwait graphic...\n","\n","[5 rows x 5 columns]"]},"metadata":{"tags":[]},"execution_count":14}]},{"cell_type":"code","metadata":{"id":"5kj2j6JLDkM_","colab_type":"code","outputId":"e3fcb392-f2a0-4834-8dde-2657596e11c4","executionInfo":{"status":"ok","timestamp":1573284425489,"user_tz":-480,"elapsed":747,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["df.shape"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(5, 5)"]},"metadata":{"tags":[]},"execution_count":15}]},{"cell_type":"code","metadata":{"id":"q6rxrqshD3M0","colab_type":"code","outputId":"590e8894-7383-47e1-8056-d5d985e4c4fa","executionInfo":{"status":"ok","timestamp":1573284469339,"user_tz":-480,"elapsed":697,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":68}},"source":["import nltk\n","nltk.download('stopwords')"],"execution_count":0,"outputs":[{"output_type":"stream","text":["[nltk_data] Downloading package stopwords to /root/nltk_data...\n","[nltk_data]   Unzipping corpora/stopwords.zip.\n"],"name":"stdout"},{"output_type":"execute_result","data":{"text/plain":["True"]},"metadata":{"tags":[]},"execution_count":18}]},{"cell_type":"code","metadata":{"id":"CbxCnYcwWugK","colab_type":"code","colab":{}},"source":["# Import nltk packages and string \n","from nltk.corpus import stopwords \n","import string\n","\n","# Define stopwords to exclude\n","stop = set(stopwords.words('english'))\n","stop.update((\"to\",\"cc\",\"subject\",\"http\",\"from\",\"sent\", \"ect\", \"u\", \"fwd\", \"www\", \"com\"))\n","\n","# Define punctuations to exclude and lemmatizer\n","exclude = set(string.punctuation)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"VVx3RuPOECLe","colab_type":"code","outputId":"7b3af125-e088-422a-ffe9-850973173122","executionInfo":{"status":"ok","timestamp":1573284584617,"user_tz":-480,"elapsed":725,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["list(stop)[:5]"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["['my', 'each', 'ourselves', 'about', 'you']"]},"metadata":{"tags":[]},"execution_count":25}]},{"cell_type":"code","metadata":{"id":"yUmiudJBEaiv","colab_type":"code","outputId":"dc4229c7-92e4-4e4e-e1a2-54695ba58870","executionInfo":{"status":"ok","timestamp":1573284591890,"user_tz":-480,"elapsed":891,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["list(exclude)[:5]"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["[')', '@', '=', '\\\\', ']']"]},"metadata":{"tags":[]},"execution_count":26}]},{"cell_type":"markdown","metadata":{"id":"aSlfQhftEe6F","colab_type":"text"},"source":["####Cleaning text data\n","\n","There are a few more steps to take before you have cleaned data, such as \"lemmatization\" of words, and stemming the verbs. \n","\n","The verbs in the email data are already stemmed, and the lemmatization is already done for you in this exercise."]},{"cell_type":"code","metadata":{"id":"guDQvSv5W1wq","colab_type":"code","colab":{}},"source":["# Import the lemmatizer from nltk\n","from nltk.stem.wordnet import WordNetLemmatizer\n","lemma = WordNetLemmatizer()\n","\n","# Define word cleaning function\n","def clean(text, stop):\n","    text = text.rstrip()\n","\t# Remove stopwords\n","    stop_free = \" \".join([word for word in text.lower().split() if ((word not in stop) and (not word.isdigit()))])\n","\t# Remove punctuations\n","    punc_free = ''.join(word for word in stop_free if word not in exclude)\n","\t# Lemmatize all words\n","    normalized = \" \".join(lemma.lemmatize(word) for word in punc_free.split())      \n","    return normalized"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"xa-BY0WDEucB","colab_type":"code","outputId":"c75f16b1-b335-4a50-ba2d-bccef479d7ef","executionInfo":{"status":"ok","timestamp":1573284677420,"user_tz":-480,"elapsed":1034,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":68}},"source":["import nltk\n","nltk.download('wordnet')"],"execution_count":0,"outputs":[{"output_type":"stream","text":["[nltk_data] Downloading package wordnet to /root/nltk_data...\n","[nltk_data]   Unzipping corpora/wordnet.zip.\n"],"name":"stdout"},{"output_type":"execute_result","data":{"text/plain":["True"]},"metadata":{"tags":[]},"execution_count":30}]},{"cell_type":"code","metadata":{"id":"9Z1dhTJHW1ue","colab_type":"code","outputId":"fbd7b401-7b7e-43ee-ec0c-1f5b4fab3b01","executionInfo":{"status":"ok","timestamp":1573284681372,"user_tz":-480,"elapsed":2598,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":54}},"source":["# Clean the emails in df and print results\n","text_clean=[]\n","for text in df['clean_content']:\n","    text_clean.append(clean(text, stop).split())    \n","print(text_clean)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["[['investools', 'advisory', 'free', 'digest', 'trusted', 'investment', 'advice', 'unsubscribe', 'free', 'newsletter', 'please', 'see', 'issue', 'fried', 'sell', 'stock', 'gain', 'month', 'km', 'rowe', 'january', 'index', 'confirms', 'bull', 'market', 'aloy', 'small', 'cap', 'advisor', 'earns', 'lbix', 'compounding', 'return', 'pine', 'tree', 'pcl', 'undervalued', 'high', 'yield', 'bank', 'put', 'customer', 'first', 'aso', 'word', 'sponsor', 'top', 'wall', 'street', 'watcher', 'ben', 'zacks', 'year', 'year', 'gain', 'moving', 'best', 'brightest', 'wall', 'street', 'big', 'money', 'machine', 'earned', 'ben', 'zacks', 'five', 'year', 'average', 'annual', 'gain', 'start', 'outperforming', 'long', 'term', 'get', 'zacks', 'latest', 'stock', 'buylist', 'free', 'day', 'trial', 'investools', 'c', 'go', 'zaks', 'mtxtu', 'zakstb', 'investools', 'advisory', 'john', 'brobst', 'investools', 'fried', 'sell', 'stock', 'lock', 'month', 'km', 'david', 'fried', 'know', 'stock', 'undervalued', 'company', 'management', 'buy', 'back', 'share', 'open', 'market', 'latest', 'triumph', 'pocketing', 'impressive', 'gain', 'three', 'short', 'month', 'selling', 'four', 'buyback', 'stock', 'include', 'gain', 'auto', 'retailer', 'automation', 'incorporated', 'gain', 'digital', 'phone', 'system', 'purveyor', 'inter', 'tel', 'intl', 'fried', 'recent', 'move', 'buy', 'kmart', 'corporation', 'km', 'beleaguered', 'discount', 'retailer', 'declared', 'bankruptcy', 'think', 'k', 'mart', 'go', 'business', 'fried', 'say', 'take', 'recovery', 'possibility', 'bought', 'share', 'another', 'fried', 'pick', 'c', 'cor', 'net', 'corporation', 'ccbl', 'provides', 'range', 'technology', 'service', 'broadband', 'network', 'today', 'telecom', 'spending', 'slowdown', 'hit', 'company', 'hard', 'net', 'sale', 'fell', 'million', 'last', 'quarter', 'caused', 'net', 'loss', 'million', 'v', 'million', 'gain', 'last', 'year', 'fried', 'cite', 'buyback', 'plan', 'million', 'restructuring', 'charge', 'proof', 'management', 'see', 'rosier', 'future', 'david', 'fried', 'advice', 'see', 'buyback', 'index', 'portfolio', 'january', 'buyback', 'letter', 'david', 'fried', 'provides', 'wealth', 'building', 'opportunity', 'company', 'repurchasing', 'stock', 'free', 'day', 'trial', 'go', 'investools', 'c', 'go', 'back', 'mtxtu', 'back', 'rowe', 'january', 'index', 'confirms', 'bull', 'market', 'aloy', 'rowe', 'say', 'january', 'index', 'confirms', 'see', 'bull', 'market', 'first', 'five', 'trading', 'day', 'provided', 'gain', 'nasdaq', 'p', 'dow', 'industrials', 'rowe', 'say', 'five', 'day', 'index', 'correctly', 'predicted', 'market', 'direction', 'year', 'since', 'four', 'exception', 'include', 'three', 'war', 'year', 'fed', 'fund', 'rate', 'doubled', 'year', 'rowe', 'maintains', 'sure', 'recommendation', 'seven', 'company', 'say', 'leading', 'market', 'one', 'alloy', 'incorporated', 'aloy', 'medium', 'company', 'direct', 'marketer', 'provides', 'content', 'community', 'commerce', 'generation', 'roughly', 'million', 'people', 'year', 'age', 'rowe', 'like', 'market', 'account', 'billion', 'disposable', 'income', 'grow', 'faster', 'overall', 'population', 'q', 'saw', 'earnings', 'increase', 'sale', 'another', 'rowe', 'pick', 'new', 'century', 'financial', 'corporation', 'ncen', 'financier', 'make', 'buy', 'sell', 'service', 'sub', 'prime', 'mortgage', 'loan', 'secured', 'first', 'mortgage', 'single', 'family', 'home', 'borrower', 'typically', 'plenty', 'equity', 'property', 'secure', 'loan', 'suffer', 'weak', 'credit', 'profile', 'high', 'debt', 'income', 'ratio', 'q', 'earnings', 'grew', 'hike', 'sale', 'rowe', 'advice', 'see', 'investment', 'opportunity', 'february', 'wall', 'street', 'digest', 'momentum', 'investor', 'donald', 'rowe', 'target', 'stock', 'mutual', 'fund', 'capable', 'generating', 'annual', 'return', 'free', 'day', 'trial', 'go', 'investools', 'c', 'go', 'wall', 'mtxtu', 'wall', 'small', 'cap', 'advisor', 'earns', 'lbix', 'major', 'index', 'suffered', 'terrible', 'year', 'richard', 'geist', 'recommendation', 'earned', 'healthy', 'list', 'many', 'reason', 'selection', 'see', 'growth', 'going', 'forward', 'include', 'extremely', 'bullish', 'monetary', 'condition', 'high', 'productivity', 'inflation', 'sight', 'yield', 'curve', 'continues', 'steepen', 'also', 'investor', 'sentiment', 'poll', 'becoming', 'bearish', 'always', 'contrary', 'indicator', 'say', 'geist', 'latest', 'recommendation', 'buy', 'share', 'leading', 'brand', 'lbix', 'company', 'canada', 'largest', 'independent', 'food', 'brand', 'management', 'company', 'expanding', 'u', 'geist', 'particularly', 'like', 'firm', 'save', 'money', 'integrated', 'distribution', 'system', 'system', 'make', 'product', 'raw', 'material', 'provides', 'packaging', 'warehousing', 'distribution', 'recent', 'financial', 'result', 'show', 'leading', 'brand', 'roll', 'fy', 'saw', 'revenue', 'grow', 'million', 'net', 'income', 'million', 'per', 'share', 'last', 'year', 'loss', 'geist', 'predicts', 'company', 'see', 'revenue', 'reach', 'million', 'million', 'yield', 'forward', 'pe', 'think', 'lbix', 'significantly', 'undervalued', 'geist', 'say', 'range', 'leading', 'brand', 'strong', 'buy', 'richard', 'geist', 'advice', 'see', 'highlighted', 'stock', 'february', 'richard', 'geist', 'strategic', 'investing', 'richard', 'geist', 'integrates', 'psychological', 'aspect', 'investing', 'methodology', 'selecting', 'small', 'company', 'stock', 'free', 'day', 'trial', 'go', 'investools', 'c', 'go', 'stin', 'mtxtu', 'stin', 'compounding', 'return', 'pine', 'tree', 'pcl', 'growing', 'tree', 'usually', 'noisy', 'business', 'catch', 'attention', 'investment', 'medium', 'good', 'business', 'say', 'dick', 'young', 'timber', 'business', 'le', 'volatile', 'capital', 'intensive', 'manufacturing', 'young', 'see', 'demand', 'timber', 'increasing', 'population', 'increase', 'note', 'average', 'return', 'timber', 'investment', 'outperformed', 'p', 'average', 'annual', 'return', 'young', 'favorite', 'timber', 'play', 'plum', 'creek', 'timber', 'pcl', 'one', 'largest', 'private', 'timberland', 'owner', 'u', 'reit', 'primary', 'goal', 'profit', 'acquiring', 'managing', 'land', 'young', 'say', 'plum', 'creek', 'timber', 'yield', 'status', 'reit', 'make', 'ideal', 'tax', 'deferred', 'account', 'another', 'young', 'timber', 'selection', 'deltic', 'timber', 'corporation', 'del', 'company', 'grows', 'harvest', 'timber', 'acre', 'arkansas', 'louisiana', 'main', 'company', 'goal', 'expand', 'timber', 'holding', 'sustainable', 'harvest', 'level', 'young', 'say', 'share', 'good', 'portfolio', 'counterweight', 'value', 'investor', 'appreciate', 'intrinsic', 'worth', 'underlying', 'real', 'natural', 'resource', 'dick', 'young', 'advice', 'see', 'investment', 'commentary', 'february', 'richard', 'young', 'intelligence', 'report', 'richard', 'young', 'us', 'buy', 'hold', 'strategy', 'mentor', 'warren', 'buffett', 'uncover', 'low', 'risk', 'high', 'reward', 'opportunity', 'free', 'day', 'trial', 'go', 'investools', 'c', 'go', 'inte', 'mtxtu', 'inte', 'undervalued', 'high', 'yield', 'bank', 'put', 'customer', 'first', 'aso', 'amsouth', 'bancorp', 'aso', 'giving', 'investor', 'healthy', 'yield', 'risk', 'involved', 'say', 'jodie', 'wei', 'investment', 'quality', 'trend', 'billion', 'asset', 'amsouth', 'one', 'largest', 'financial', 'institution', 'south', 'office', 'credit', 'bank', 'success', 'putting', 'customer', 'first', 'wei', 'like', 'amsouth', 'us', 'new', 'technology', 'save', 'money', 'streamlining', 'operation', 'note', 'amsouth', 'ranked', 'number', 'six', 'eweek', 'fast', 'track', 'list', 'company', 'deploy', 'cutting', 'edge', 'technology', 'throughout', 'operation', 'number', 'merrill', 'lynch', 'financial', 'service', 'firm', 'placed', 'higher', 'also', 'amsouth', 'internet', 'banking', 'group', 'quadrupled', 'customer', 'base', 'last', 'year', 'wei', 'say', 'aso', 'share', 'undervalued', 'stock', 'selling', 'near', 'yield', 'wei', 'see', 'upside', 'potential', 'dividend', 'risen', 'annually', 'past', 'year', 'buyback', 'plan', 'million', 'share', 'authorized', 'september', 'stock', 'pe', 'reasonable', 'x', 'priced', 'yield', 'aso', 'undervalued', 'buy', 'considered', 'wei', 'say', 'jodie', 'wei', 'advice', 'see', 'investment', 'spotlight', 'january', 'income', 'digest', 'digest', 'excerpt', 'investment', 'publication', 'highlight', 'weather', 'income', 'oriented', 'opportunity', 'uncovered', 'top', 'mind', 'wall', 'street', 'free', 'day', 'trial', 'go', 'investools', 'c', 'go', 'indi', 'mtxtu', 'indi', 'word', 'sponsor', 'new', 'report', 'top', 'pick', 'despite', 'slumping', 'economy', 'shaky', 'stock', 'market', 'frank', 'curzio', 'bull', 'eye', 'pick', 'gained', 'whopping', 'curzio', 'selected', 'stock', 'incredible', 'potential', 'get', 'red', 'hot', 'pick', 'today', 'click', 'investools', 'c', 'go', 'fxcp', 'fxcp', 'mtxtu', 'disclaimer', 'investools', 'advisory', 'published', 'solely', 'informational', 'purpose', 'solicit', 'offer', 'buy', 'sell', 'stock', 'mutual', 'fund', 'security', 'attempt', 'claim', 'complete', 'description', 'security', 'market', 'development', 'referred', 'material', 'expression', 'opinion', 'change', 'without', 'notice', 'information', 'obtained', 'internal', 'external', 'source', 'investools', 'considers', 'reliable', 'investools', 'independently', 'verified', 'information', 'investools', 'guarantee', 'accurate', 'complete', 'investools', 'undertake', 'advise', 'anyone', 'investools', 'employee', 'officer', 'director', 'may', 'time', 'time', 'position', 'security', 'mentioned', 'may', 'sell', 'buy', 'security', 'remove', 'free', 'email', 'list', 'removed', 'email', 'distribution', 'list', 'free', 'investools', 'advisory', 'update', 'simply', 'click', 'link', 'hit', 'send', 'email', 'launched', 'copy', 'paste', 'email', 'address', 'new', 'outgoing', 'email', 'message', 'hit', 'send', 'email', 'launched', 'mailto', 'bonnie', 'investools', 'important', 'automated', 'system', 'cancel', 'paid', 'newsletter', 'service', 'subscription', 'investools', 'tried', 'unsubscribing', 'past', 'believe', 'received', 'message', 'error', 'please', 'send', 'email', 'mailto', 'itfeedback', 'investools', 'voice', 'concern', 'removed', 'list', 'paid', 'subscription', 'information', 'question', 'investools', 'service', 'paid', 'subscription', 'contact', 'investools', 'customer', 'service', 'center', 'investools', 'cgi', 'bin', 'help', 'pl', 'info', 'pr', 'faq', 'html'], ['forwarded', 'richard', 'b', 'sander', 'hou', 'pm', 'justin', 'boyd', 'richard', 'b', 'sander', 'hou', 'mg', 'plc', 'rudolph', 'wolfe', 'richard', 'please', 'see', 'justin', 'forwarded', 'justin', 'boyd', 'lon', 'justin', 'boyd', 'philippa', 'broom', 'glegal', 'enron', 'mg', 'plc', 'rudolph', 'wolfe', 'many', 'thanks', 'enron', 'capital', 'trade', 'resource', 'corp', 'philippa', 'broom', 'glegal', 'justin', 'boyd', 'enron', 'james', 'greig', 'glegal', 'mg', 'plc', 'rudolph', 'wolfe', 'justin', 'uk', 'agreement', 'metallgesellschaft', 'limited', 'mgl', 'rudolf', 'wolff', 'group', 'limited', 'rwg', 'dated', 'june', 'mgl', 'purchased', 'certain', 'lme', 'warrant', 'metal', 'rwg', 'mgl', 'sold', 'certain', 'lme', 'future', 'rwg', 'mgl', 'agreed', 'make', 'certain', 'payment', 'rwg', 'use', 'rwg', 'london', 'hamburg', 'new', 'york', 'office', 'month', 'following', 'closing', 'party', 'agreed', 'process', 'transfer', 'rwg', 'customer', 'mgl', 'mgl', 'agreed', 'make', 'offer', 'employment', 'certain', 'identified', 'employee', 'director', 'rwg', 'group', 'rwg', 'assigned', 'mgl', 'rudolf', 'wolff', 'trademark', 'including', 'company', 'name', 'logo', 'domain', 'name', 'website', 'rwg', 'gave', 'mgl', 'two', 'year', 'non', 'compete', 'covenant', 'u', 'agreement', 'mg', 'metal', 'commodity', 'co', 'limited', 'mgm', 'rudolf', 'wolff', 'metal', 'co', 'inc', 'rwmci', 'dated', 'june', 'mgm', 'purchased', 'certain', 'u', 'non', 'u', 'metal', 'inventory', 'related', 'future', 'contract', 'hedge', 'rwmci', 'rwmci', 'assigned', 'certain', 'forward', 'physical', 'contract', 'mgm', 'mgm', 'agreed', 'make', 'offer', 'employment', 'employee', 'rwmci', 'neither', 'transaction', 'involved', 'sale', 'share', 'rudolf', 'wolff', 'group', 'mg', 'need', 'information', 'please', 'hesitate', 'contact', 'regard', 'philippa', 'philippa', 'broom', 'date', 'justin', 'boyd', 'enron', 'mg', 'plc', 'rudolph', 'wolfe', 'philippa', 'please', 'would', 'prepare', 'brief', 'description', 'underlying', 'rw', 'deal', 'uk', 'u', 'jane', 'allen', 'purpose', 'make', 'short', 'sweet', 'thanks', 'justin', 'forwarded', 'justin', 'boyd', 'lon', 'jeanie', 'slone', 'justin', 'boyd', 'lon', 'madeline', 'fox', 'lon', 'mg', 'plc', 'rudolph', 'wolfe', 'justin', 'please', 'see', 'attached', 'email', 'jane', 'allen', 'houston', 'could', 'provide', 'diagram', 'requested', 'please', 'respond', 'madeline', 'fox', 'thanks', 'best', 'regard', 'jeanie', 'forwarded', 'jeanie', 'slone', 'lon', 'madeline', 'fox', 'jeanie', 'slone', 'lon', 'mg', 'plc', 'rudolph', 'wolfe', 'jeanie', 'want', 'burden', 'last', 'week', 'would', 'easy', 'draw', 'thanks', 'mad', 'forwarded', 'madeline', 'fox', 'lon', 'jane', 'allen', 'hoyt', 'thomas', 'na', 'enron', 'enron', 'madeline', 'fox', 'lon', 'margaret', 'daffin', 'hou', 'melissa', 'laing', 'lon', 'mg', 'plc', 'rudolph', 'wolfe', 'spoken', 'alice', 'gruber', 'immigration', 'attorney', 'order', 'determine', 'get', 'l', 'transferred', 'need', 'basic', 'diagram', 'transaction', 'company', 'example', 'acquired', 'percentage', 'well', 'whether', 'transaction', 'stock', 'based', 'asset', 'acquisition', 'merger', 'please', 'send', 'directly', 'work', 'alice', 'jane', 'j', 'allen', 'enron', 'corp', 'global', 'employee', 'service', 'phone', 'fax', 'forwarded', 'jane', 'allen', 'hou', 'margaret', 'daffin', 'alice', 'gruber', 'agruber', 'tindallfoster', 'enron', 'susan', 'king', 'sking', 'tindallfoster', 'enron', 'jane', 'allen', 'hou', 'melissa', 'laing', 'lon', 'madeline', 'fox', 'lon', 'hoyt', 'thomas', 'na', 'enron', 'enron', 'mg', 'plc', 'rudolph', 'wolfe', 'alice', 'reference', 'following', 'enron', 'recently', 'made', 'asset', 'acquisition', 'rudolf', 'wolf', 'received', 'following', 'note', 'london', 'need', 'help', 'sorting', 'paul', 'rudolf', 'wolf', 'mg', 'owned', 'company', 'paul', 'uk', 'national', 'l', 'visa', 'obtained', 'rudolf', 'wolf', 'company', 'transitioning', 'shortly', 'enron', 'contract', 'sure', 'need', 'immigration', 'perspective', 'transition', 'l', 'visa', 'could', 'let', 'know', 'information', 'lawyer', 'need', 'get', 'ball', 'rolling', 'please', 'let', 'know', 'whether', 'possible', 'bring', 'paul', 'fingster', 'l', 'given', 'fact', 'enron', 'stock', 'transfer', 'many', 'thanks', 'margaret', 'forwarded', 'margaret', 'daffin', 'hou', 'jane', 'allen', 'margaret', 'daffin', 'hou', 'mg', 'plc', 'rudolph', 'wolfe', 'jane', 'j', 'allen', 'enron', 'corp', 'global', 'employee', 'service', 'phone', 'fax', 'forwarded', 'jane', 'allen', 'hou', 'jane', 'allen', 'hoyt', 'thomas', 'na', 'enron', 'enron', 'melissa', 'laing', 'lon', 'enron', 'pm', 'mg', 'plc', 'rudolph', 'wolfe', 'document', 'link', 'jane', 'allen', 'sorry', 'vacation', 'call', 'ingrid', 'wanted', 'see', 'found', 'anything', 'trip', 'ny', 'mg', 'mcc', 'immigration', 'attorney', 'assist', 'visa', 'filing', 'able', 'coach', 'nervous', 'coaching', 'law', 'firm', 'handle', 'visa', 'issue', 'case', 'may', 'different', 'law', 'firm', 'speak', 'individual', 'determine', 'best', 'route', 'take', 'complete', 'paperwork', 'sign', 'jane', 'j', 'allen', 'enron', 'corp', 'global', 'employee', 'service', 'phone', 'fax', 'hoyt', 'thomas', 'enron', 'jane', 'allen', 'hou', 'melissa', 'laing', 'lon', 'mg', 'plc', 'rudolph', 'wolfe', 'pm', 'hi', 'jane', 'talked', 'ingrid', 'krinke', 'today', 'ingrid', 'hr', 'person', 'mg', 'mcc', 'new', 'york', 'branch', 'mg', 'plc', 'mg', 'plc', 'acquired', 'rudolph', 'wolff', 'another', 'british', 'company', 'rudolph', 'wolfe', 'two', 'h', 'one', 'l', 'visa', 'employee', 'turn', 'mg', 'plc', 'asset', 'acquisition', 'rudolph', 'wolff', 'stock', 'transfer', 'ingrid', 'immigration', 'attorney', 'tell', 'cannot', 'transfer', 'l', 'asset', 'transfer', 'ingrid', 'starting', 'process', 'transferring', 'h', 'rw', 'mg', 'know', 'much', 'ingrid', 'know', 'visa', 'would', 'probably', 'good', 'guided', 'something', 'impact', 'u', 'unnecessary', 'ingrid', 'thick', 'file', 'immigration', 'stuff', 'folk', 'know', 'send', 'u', 'could', 'give', 'call', 'talk', 'situation', 'concerned', 'asking', 'right', 'question', 'may', 'garble', 'info', 'pas', 'ingrid', 'told', 'might', 'call', 'closing', 'forecast', 'sept', 'may', 'ny', 'either', 'monday', 'following', 'monday', 'collecting', 'data', 'would', 'happy', 'sit', 'tell', 'know', 'join', 'call', 'ingrid', 'please', 'ring', 'thanks', 'bunch', 'internet', 'email', 'confidentiality', 'footer', 'garretts', 'member', 'andersen', 'worldwide', 'international', 'network', 'law', 'firm', 'list', 'name', 'partner', 'professional', 'qualification', 'open', 'inspection', 'strand', 'london', 'wc', 'r', 'nn', 'principal', 'place', 'business', 'partner', 'either', 'solicitor', 'registered', 'foreign', 'lawyer', 'associate', 'firm', 'scotland', 'dundas', 'wilson', 'privileged', 'confidential', 'information', 'may', 'contained', 'message', 'addressee', 'indicated', 'message', 'responsible', 'delivery', 'message', 'person', 'may', 'copy', 'deliver', 'message', 'anyone', 'case', 'destroy', 'message', 'kindly', 'notify', 'sender', 'reply', 'email', 'please', 'advise', 'immediately', 'employer', 'consent', 'internet', 'email', 'message', 'kind', 'opinion', 'conclusion', 'information', 'message', 'relate', 'official', 'business', 'firm', 'shall', 'understood', 'neither', 'given', 'endorsed'], ['hey', 'wearing', 'target', 'purple', 'shirt', 'today', 'mine', 'want', 'look', 'silly', 'original', 'message', 'baumbach', 'david', 'tuesday', 'october', 'love', 'phillip', 'league', 'due', 'team', 'look', 'good', 'paper', 'come', 'play', 'rc', 'going', 'call', 'original', 'message', 'love', 'phillip', 'tuesday', 'october', 'baumbach', 'david', 'league', 'due', 'much', 'nice', 'machine', 'original', 'message', 'baumbach', 'david', 'tuesday', 'october', 'bass', 'eric', 'ames', 'chuck', 'bucalo', 'harry', 'bucalo', 'harry', 'hull', 'bryan', 'jason', 'bass', 'e', 'mail', 'lenhart', 'matthew', 'love', 'phillip', 'mathews', 'reagan', 'mill', 'bruce', 'rabon', 'chance', 'winfree', 'neal', 'league', 'due', 'way', 'looking', 'gave', 'money', 'away', 'anyway', 'original', 'message', 'bass', 'eric', 'tuesday', 'october', 'ames', 'chuck', 'baumbach', 'david', 'bucalo', 'harry', 'bucalo', 'harry', 'hull', 'bryan', 'jason', 'bass', 'e', 'mail', 'lenhart', 'matthew', 'love', 'phillip', 'mathews', 'reagan', 'mill', 'bruce', 'rabon', 'chance', 'winfree', 'neal', 'league', 'due', 'good', 'news', 'bad', 'news', 'good', 'news', 'invested', 'league', 'due', 'stock', 'market', 'last', 'week', 'bad', 'news', 'invested', 'enron', 'stock', 'needle', 'say', 'going', 'default', 'year', 'payouts', 'sorry', 'guy'], ['leslie', 'milosevich', 'santa', 'clara', 'avenue', 'alameda', 'ca', 'leslie', 'milosevich', 'kp', 'org', 'mr', 'ken', 'lay', 'writing', 'urge', 'donate', 'million', 'dollar', 'made', 'selling', 'enron', 'stock', 'company', 'declared', 'bankruptcy', 'fund', 'enron', 'employee', 'transition', 'fund', 'reach', 'benefit', 'company', 'employee', 'lost', 'retirement', 'saving', 'provide', 'relief', 'low', 'income', 'consumer', 'california', 'afford', 'pay', 'energy', 'bill', 'enron', 'made', 'million', 'pocketbook', 'california', 'consumer', 'effort', 'employee', 'indeed', 'netted', 'well', 'million', 'many', 'enron', 'employee', 'financially', 'devastated', 'company', 'declared', 'bankruptcy', 'retirement', 'plan', 'wiped', 'enron', 'made', 'astronomical', 'profit', 'california', 'energy', 'crisis', 'last', 'year', 'result', 'thousand', 'consumer', 'unable', 'pay', 'basic', 'energy', 'bill', 'largest', 'utility', 'state', 'bankrupt', 'new', 'york', 'time', 'reported', 'sold', 'million', 'worth', 'enron', 'stock', 'aggressively', 'urging', 'company', 'employee', 'keep', 'buying', 'please', 'donate', 'money', 'fund', 'set', 'help', 'repair', 'life', 'american', 'hurt', 'enron', 'underhanded', 'dealing', 'sincerely', 'leslie', 'milosevich'], ['rini', 'twait', 'e', 'th', 'ave', 'longmont', 'co', 'rtwait', 'graphicaljazz', 'mr', 'ken', 'lay', 'writing', 'urge', 'donate', 'million', 'dollar', 'made', 'selling', 'enron', 'stock', 'company', 'declared', 'bankruptcy', 'fund', 'enron', 'employee', 'transition', 'fund', 'reach', 'benefit', 'company', 'employee', 'lost', 'retirement', 'saving', 'provide', 'relief', 'low', 'income', 'consumer', 'california', 'afford', 'pay', 'energy', 'bill', 'enron', 'made', 'million', 'pocketbook', 'california', 'consumer', 'effort', 'employee', 'indeed', 'netted', 'well', 'million', 'many', 'enron', 'employee', 'financially', 'devastated', 'company', 'declared', 'bankruptcy', 'retirement', 'plan', 'wiped', 'enron', 'made', 'astronomical', 'profit', 'california', 'energy', 'crisis', 'last', 'year', 'result', 'thousand', 'consumer', 'unable', 'pay', 'basic', 'energy', 'bill', 'largest', 'utility', 'state', 'bankrupt', 'new', 'york', 'time', 'reported', 'sold', 'million', 'worth', 'enron', 'stock', 'aggressively', 'urging', 'company', 'employee', 'keep', 'buying', 'please', 'donate', 'money', 'fund', 'set', 'help', 'repair', 'life', 'american', 'hurt', 'enron', 'underhanded', 'dealing', 'sincerely', 'rini', 'twait']]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"cVzcQHV6W9xt","colab_type":"text"},"source":["###Topic modeling on fraud\n"]},{"cell_type":"markdown","metadata":{"id":"1MOQZ9HYE0Iq","colab_type":"text"},"source":["####Create dictionary and corpus\n","\n","In order to run an LDA topic model, you first need to define your dictionary and corpus first, as those need to go into the model. "]},{"cell_type":"code","metadata":{"id":"CCpCpKmgWueY","colab_type":"code","outputId":"4fee7aec-8114-4ec1-cef4-04aa08cecebb","executionInfo":{"status":"ok","timestamp":1573284725938,"user_tz":-480,"elapsed":751,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":71}},"source":["# Import the packages\n","import gensim\n","from gensim import corpora\n","\n","# Define the dictionary\n","dictionary = corpora.Dictionary(text_clean)\n","\n","# Define the corpus\n","corpus = [dictionary.doc2bow(text) for text in text_clean]\n","\n","# Print corpus and dictionary\n","print(dictionary)\n","print(corpus)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Dictionary(924 unique tokens: ['account', 'accurate', 'acquiring', 'acre', 'address']...)\n","[[(0, 2), (1, 1), (2, 1), (3, 1), (4, 1), (5, 6), (6, 1), (7, 2), (8, 4), (9, 1), (10, 1), (11, 3), (12, 2), (13, 1), (14, 5), (15, 3), (16, 1), (17, 3), (18, 1), (19, 1), (20, 1), (21, 5), (22, 1), (23, 1), (24, 1), (25, 1), (26, 1), (27, 1), (28, 1), (29, 1), (30, 3), (31, 3), (32, 1), (33, 3), (34, 1), (35, 1), (36, 1), (37, 1), (38, 1), (39, 1), (40, 1), (41, 2), (42, 1), (43, 1), (44, 2), (45, 1), (46, 1), (47, 1), (48, 1), (49, 4), (50, 1), (51, 1), (52, 1), (53, 1), (54, 1), (55, 4), (56, 1), (57, 4), (58, 9), (59, 5), (60, 1), (61, 8), (62, 1), (63, 1), (64, 2), (65, 1), (66, 1), (67, 1), (68, 1), (69, 1), (70, 1), (71, 1), (72, 1), (73, 1), (74, 1), (75, 1), (76, 1), (77, 2), (78, 1), (79, 1), (80, 1), (81, 12), (82, 2), (83, 2), (84, 1), (85, 1), (86, 3), (87, 1), (88, 1), (89, 1), (90, 1), (91, 1), (92, 1), (93, 1), (94, 1), (95, 4), (96, 1), (97, 1), (98, 2), (99, 2), (100, 1), (101, 2), (102, 5), (103, 1), (104, 3), (105, 8), (106, 1), (107, 1), (108, 1), (109, 1), (110, 1), (111, 1), (112, 1), (113, 1), (114, 1), (115, 1), (116, 2), (117, 4), (118, 1), (119, 1), (120, 1), (121, 1), (122, 1), (123, 1), (124, 1), (125, 3), (126, 1), (127, 1), (128, 1), (129, 1), (130, 2), (131, 2), (132, 2), (133, 1), (134, 1), (135, 7), (136, 1), (137, 1), (138, 1), (139, 1), (140, 1), (141, 1), (142, 1), (143, 1), (144, 1), (145, 1), (146, 1), (147, 1), (148, 1), (149, 1), (150, 1), (151, 1), (152, 1), (153, 3), (154, 1), (155, 1), (156, 4), (157, 1), (158, 2), (159, 5), (160, 3), (161, 1), (162, 2), (163, 2), (164, 1), (165, 10), (166, 9), (167, 3), (168, 1), (169, 2), (170, 1), (171, 8), (172, 1), (173, 8), (174, 1), (175, 1), (176, 2), (177, 1), (178, 13), (179, 2), (180, 1), (181, 2), (182, 1), (183, 1), (184, 2), (185, 1), (186, 1), (187, 1), (188, 1), (189, 1), (190, 2), (191, 2), (192, 1), (193, 5), (194, 1), (195, 1), (196, 1), (197, 1), (198, 3), (199, 1), (200, 1), (201, 1), (202, 1), (203, 1), (204, 1), (205, 1), (206, 1), (207, 3), (208, 5), (209, 2), (210, 2), (211, 1), (212, 1), (213, 1), (214, 1), (215, 6), (216, 2), (217, 1), (218, 1), (219, 1), (220, 1), (221, 3), (222, 1), (223, 1), (224, 2), (225, 1), (226, 1), (227, 1), (228, 1), (229, 1), (230, 1), (231, 1), (232, 1), (233, 1), (234, 2), (235, 8), (236, 23), (237, 4), (238, 1), (239, 1), (240, 1), (241, 5), (242, 2), (243, 1), (244, 1), (245, 3), (246, 1), (247, 1), (248, 1), (249, 3), (250, 4), (251, 3), (252, 2), (253, 4), (254, 1), (255, 4), (256, 1), (257, 1), (258, 3), (259, 1), (260, 5), (261, 2), (262, 1), (263, 1), (264, 2), (265, 1), (266, 1), (267, 1), (268, 1), (269, 2), (270, 1), (271, 1), (272, 1), (273, 3), (274, 3), (275, 1), (276, 1), (277, 1), (278, 9), (279, 1), (280, 1), (281, 2), (282, 2), (283, 2), (284, 1), (285, 1), (286, 1), (287, 2), (288, 1), (289, 10), (290, 1), (291, 1), (292, 1), (293, 3), (294, 3), (295, 2), (296, 1), (297, 1), (298, 7), (299, 2), (300, 1), (301, 1), (302, 1), (303, 1), (304, 4), (305, 1), (306, 4), (307, 2), (308, 1), (309, 2), (310, 1), (311, 2), (312, 1), (313, 1), (314, 1), (315, 1), (316, 3), (317, 1), (318, 2), (319, 1), (320, 4), (321, 1), (322, 1), (323, 1), (324, 1), (325, 1), (326, 1), (327, 2), (328, 1), (329, 3), (330, 1), (331, 2), (332, 1), (333, 3), (334, 2), (335, 1), (336, 1), (337, 1), (338, 5), (339, 2), (340, 1), (341, 1), (342, 2), (343, 1), (344, 2), (345, 1), (346, 2), (347, 1), (348, 1), (349, 2), (350, 2), (351, 1), (352, 1), (353, 2), (354, 1), (355, 1), (356, 1), (357, 1), (358, 1), (359, 1), (360, 1), (361, 1), (362, 1), (363, 1), (364, 1), (365, 1), (366, 1), (367, 1), (368, 4), (369, 1), (370, 1), (371, 1), (372, 1), (373, 1), (374, 2), (375, 1), (376, 2), (377, 1), (378, 1), (379, 1), (380, 1), (381, 2), (382, 1), (383, 1), (384, 1), (385, 1), (386, 1), (387, 1), (388, 1), (389, 1), (390, 1), (391, 2), (392, 3), (393, 1), (394, 1), (395, 1), (396, 2), (397, 1), (398, 1), (399, 2), (400, 2), (401, 1), (402, 1), (403, 1), (404, 1), (405, 2), (406, 5), (407, 2), (408, 1), (409, 6), (410, 1), (411, 2), (412, 1), (413, 1), (414, 1), (415, 9), (416, 3), (417, 2), (418, 2), (419, 12), (420, 1), (421, 1), (422, 4), (423, 12), (424, 1), (425, 1), (426, 2), (427, 5), (428, 2), (429, 3), (430, 1), (431, 1), (432, 6), (433, 1), (434, 1), (435, 7), (436, 1), (437, 1), (438, 1), (439, 1), (440, 1), (441, 1), (442, 1), (443, 1), (444, 1), (445, 1), (446, 3), (447, 1), (448, 1), (449, 1), (450, 1), (451, 1), (452, 2), (453, 1), (454, 1), (455, 1), (456, 1), (457, 2), (458, 14), (459, 1), (460, 1), (461, 1), (462, 4), (463, 1), (464, 1), (465, 3), (466, 1), (467, 1), (468, 1), (469, 1), (470, 1), (471, 4), (472, 1), (473, 1), (474, 1), (475, 3), (476, 1), (477, 1), (478, 1), (479, 1), (480, 2), (481, 2), (482, 1), (483, 10), (484, 1), (485, 2), (486, 2), (487, 3), (488, 1), (489, 1), (490, 3), (491, 1), (492, 6), (493, 1), (494, 1), (495, 1), (496, 1), (497, 2), (498, 1), (499, 1), (500, 1), (501, 1), (502, 6), (503, 1), (504, 1), (505, 1), (506, 1), (507, 2), (508, 1), (509, 1), (510, 1), (511, 1), (512, 1), (513, 1), (514, 6), (515, 1), (516, 1), (517, 1), (518, 1), (519, 1), (520, 1), (521, 1), (522, 6), (523, 1), (524, 1), (525, 2), (526, 1), (527, 1), (528, 12), (529, 8), (530, 9), (531, 3), (532, 1), (533, 1)], [(6, 1), (17, 1), (18, 1), (23, 4), (42, 2), (57, 2), (66, 1), (81, 5), (82, 1), (89, 1), (93, 1), (102, 1), (113, 1), (121, 1), (135, 4), (136, 6), (158, 5), (162, 1), (168, 2), (176, 2), (181, 1), (183, 3), (192, 1), (220, 1), (221, 4), (231, 2), (238, 1), (239, 1), (247, 6), (250, 1), (259, 1), (260, 1), (273, 4), (277, 2), (282, 5), (287, 7), (294, 1), (305, 1), (306, 2), (309, 1), (312, 1), (313, 2), (314, 1), (316, 1), (317, 1), (319, 1), (337, 3), (344, 9), (372, 1), (380, 1), (390, 1), (402, 1), (409, 3), (416, 1), (423, 3), (429, 2), (432, 3), (435, 1), (436, 1), (458, 3), (469, 1), (472, 1), (486, 1), (497, 6), (500, 1), (528, 1), (534, 1), (535, 2), (536, 3), (537, 1), (538, 4), (539, 2), (540, 1), (541, 4), (542, 13), (543, 1), (544, 1), (545, 1), (546, 2), (547, 1), (548, 1), (549, 1), (550, 3), (551, 2), (552, 1), (553, 1), (554, 1), (555, 7), (556, 1), (557, 1), (558, 1), (559, 1), (560, 3), (561, 1), (562, 1), (563, 4), (564, 1), (565, 2), (566, 6), (567, 2), (568, 2), (569, 1), (570, 1), (571, 1), (572, 1), (573, 1), (574, 1), (575, 1), (576, 1), (577, 1), (578, 1), (579, 1), (580, 3), (581, 4), (582, 3), (583, 1), (584, 4), (585, 1), (586, 1), (587, 2), (588, 1), (589, 1), (590, 1), (591, 1), (592, 2), (593, 2), (594, 1), (595, 1), (596, 1), (597, 1), (598, 1), (599, 1), (600, 1), (601, 2), (602, 1), (603, 2), (604, 1), (605, 20), (606, 1), (607, 1), (608, 3), (609, 1), (610, 1), (611, 1), (612, 1), (613, 4), (614, 1), (615, 1), (616, 1), (617, 8), (618, 1), (619, 6), (620, 1), (621, 1), (622, 1), (623, 1), (624, 2), (625, 3), (626, 3), (627, 1), (628, 2), (629, 1), (630, 2), (631, 1), (632, 1), (633, 1), (634, 1), (635, 1), (636, 1), (637, 9), (638, 1), (639, 4), (640, 1), (641, 1), (642, 1), (643, 5), (644, 1), (645, 1), (646, 1), (647, 1), (648, 1), (649, 9), (650, 1), (651, 1), (652, 1), (653, 3), (654, 1), (655, 14), (656, 5), (657, 1), (658, 2), (659, 11), (660, 1), (661, 1), (662, 1), (663, 1), (664, 6), (665, 4), (666, 3), (667, 2), (668, 2), (669, 3), (670, 2), (671, 1), (672, 13), (673, 3), (674, 1), (675, 1), (676, 6), (677, 5), (678, 2), (679, 4), (680, 1), (681, 1), (682, 4), (683, 1), (684, 20), (685, 8), (686, 4), (687, 1), (688, 2), (689, 1), (690, 3), (691, 3), (692, 1), (693, 5), (694, 2), (695, 1), (696, 1), (697, 2), (698, 1), (699, 2), (700, 1), (701, 1), (702, 1), (703, 1), (704, 2), (705, 1), (706, 1), (707, 3), (708, 1), (709, 1), (710, 2), (711, 1), (712, 5), (713, 1), (714, 1), (715, 14), (716, 3), (717, 1), (718, 1), (719, 1), (720, 1), (721, 1), (722, 2), (723, 1), (724, 1), (725, 2), (726, 1), (727, 1), (728, 1), (729, 1), (730, 2), (731, 1), (732, 1), (733, 1), (734, 1), (735, 1), (736, 1), (737, 1), (738, 1), (739, 1), (740, 1), (741, 1), (742, 7), (743, 14), (744, 2), (745, 9), (746, 4), (747, 2), (748, 1), (749, 1), (750, 1), (751, 1), (752, 1), (753, 1), (754, 1), (755, 1), (756, 1), (757, 3), (758, 1), (759, 1), (760, 1), (761, 1), (762, 1), (763, 1), (764, 1), (765, 1), (766, 1), (767, 1), (768, 1), (769, 1), (770, 1), (771, 1), (772, 2), (773, 6), (774, 1), (775, 4), (776, 2), (777, 1), (778, 1), (779, 1), (780, 3), (781, 5), (782, 1), (783, 1), (784, 1), (785, 1), (786, 1), (787, 1), (788, 2), (789, 3), (790, 1), (791, 1), (792, 1), (793, 1), (794, 6), (795, 1), (796, 1), (797, 1), (798, 1), (799, 1), (800, 1), (801, 1), (802, 2), (803, 1), (804, 3), (805, 12), (806, 6), (807, 1), (808, 1), (809, 4), (810, 2)], [(104, 4), (180, 2), (181, 3), (250, 1), (268, 1), (278, 1), (287, 4), (293, 1), (343, 1), (419, 1), (458, 2), (473, 1), (486, 1), (528, 1), (563, 1), (605, 1), (622, 1), (689, 1), (761, 1), (795, 1), (800, 1), (811, 2), (812, 1), (813, 1), (814, 2), (815, 4), (816, 4), (817, 2), (818, 2), (819, 4), (820, 2), (821, 2), (822, 1), (823, 1), (824, 5), (825, 2), (826, 2), (827, 1), (828, 4), (829, 1), (830, 2), (831, 2), (832, 2), (833, 5), (834, 2), (835, 2), (836, 1), (837, 4), (838, 2), (839, 2), (840, 2), (841, 2), (842, 1), (843, 2), (844, 1), (845, 4), (846, 1), (847, 4), (848, 4), (849, 1), (850, 1), (851, 4), (852, 1), (853, 2), (854, 1), (855, 2), (856, 1), (857, 1), (858, 1), (859, 4), (860, 1), (861, 1), (862, 2)], [(35, 2), (81, 4), (107, 2), (136, 5), (167, 3), (192, 1), (208, 1), (249, 1), (250, 1), (266, 1), (277, 1), (289, 4), (293, 1), (306, 1), (342, 1), (344, 1), (364, 1), (386, 1), (404, 1), (428, 1), (458, 2), (485, 1), (526, 1), (528, 1), (554, 1), (605, 7), (675, 3), (724, 1), (758, 1), (784, 1), (801, 1), (810, 1), (863, 1), (864, 1), (865, 1), (866, 1), (867, 1), (868, 1), (869, 1), (870, 1), (871, 2), (872, 1), (873, 1), (874, 3), (875, 1), (876, 3), (877, 1), (878, 1), (879, 1), (880, 1), (881, 2), (882, 1), (883, 3), (884, 1), (885, 1), (886, 1), (887, 1), (888, 1), (889, 1), (890, 1), (891, 3), (892, 1), (893, 1), (894, 3), (895, 1), (896, 1), (897, 1), (898, 2), (899, 1), (900, 1), (901, 1), (902, 1), (903, 2), (904, 1), (905, 1), (906, 1), (907, 1), (908, 1), (909, 1), (910, 1), (911, 1), (912, 1), (913, 1), (914, 1), (915, 1), (916, 1)], [(35, 2), (81, 4), (107, 2), (136, 5), (167, 3), (192, 1), (208, 1), (249, 1), (250, 1), (266, 1), (277, 1), (289, 4), (293, 1), (306, 1), (342, 1), (344, 1), (364, 1), (386, 1), (404, 1), (428, 1), (458, 2), (485, 1), (526, 1), (528, 1), (554, 1), (568, 1), (605, 7), (675, 3), (724, 1), (758, 1), (784, 1), (801, 1), (810, 1), (825, 1), (863, 1), (864, 1), (866, 1), (867, 1), (869, 1), (870, 1), (871, 2), (872, 1), (874, 3), (876, 3), (877, 1), (878, 1), (879, 1), (880, 1), (881, 2), (882, 1), (883, 3), (884, 1), (885, 1), (886, 1), (887, 1), (888, 1), (890, 1), (892, 1), (893, 1), (895, 1), (896, 1), (898, 2), (899, 1), (900, 1), (901, 1), (902, 1), (903, 2), (905, 1), (906, 1), (907, 1), (908, 1), (909, 1), (910, 1), (911, 1), (912, 1), (913, 1), (914, 1), (915, 1), (916, 1), (917, 1), (918, 1), (919, 1), (920, 2), (921, 1), (922, 1), (923, 2)]]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"aypa2cPIFA_8","colab_type":"text"},"source":["####LDA model\n","\n","- Build the LDA model from gensim models, by inserting the corpus and dictionary.\n","- Save the 5 topics by running print topics on the model results, and select the top 5 words."]},{"cell_type":"code","metadata":{"id":"iTEom76WVt-r","colab_type":"code","outputId":"e0b8ec14-da9b-449a-b2ae-4127064a59d9","executionInfo":{"status":"ok","timestamp":1573284780281,"user_tz":-480,"elapsed":741,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":102}},"source":["# Define the LDA model\n","ldamodel = gensim.models.ldamodel.LdaModel(corpus, num_topics=5, id2word=dictionary, passes=5)\n","\n","# Save the topics and top 5 words\n","topics = ldamodel.print_topics(num_words=5)\n","\n","# Print the results\n","for topic in topics:\n","    print(topic)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["(0, '0.003*\"enron\" + 0.002*\"mg\" + 0.002*\"investools\" + 0.002*\"company\" + 0.002*\"stock\"')\n","(1, '0.025*\"enron\" + 0.017*\"employee\" + 0.014*\"company\" + 0.014*\"million\" + 0.011*\"stock\"')\n","(2, '0.002*\"enron\" + 0.002*\"rudolph\" + 0.002*\"mg\" + 0.002*\"jane\" + 0.002*\"investools\"')\n","(3, '0.019*\"investools\" + 0.011*\"stock\" + 0.011*\"go\" + 0.010*\"company\" + 0.010*\"say\"')\n","(4, '0.021*\"enron\" + 0.021*\"mg\" + 0.015*\"plc\" + 0.014*\"rudolph\" + 0.014*\"jane\"')\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"4Uz8P496XIxF","colab_type":"text"},"source":["###Flagging fraud based on topics\n"]},{"cell_type":"code","metadata":{"id":"uHgCZpDuFZgE","colab_type":"code","colab":{}},"source":["def get_topic_details(ldamodel, corpus):\n","    topic_details_df = pd.DataFrame()\n","    for i, row in enumerate(ldamodel[corpus]):\n","        row = sorted(row, key=lambda x: (x[1]), reverse=True)\n","        for j, (topic_num, prop_topic) in enumerate(row):\n","            if j == 0:  # => dominant topic\n","                wp = ldamodel.show_topic(topic_num)\n","                topic_keywords = \", \".join([word for word, prop in wp])\n","                topic_details_df = topic_details_df.append(pd.Series([int(topic_num), round(prop_topic,4), topic_keywords]), ignore_index=True)\n","    topic_details_df.columns = ['Dominant_Topic', '% Score', 'Topic_Keywords']\n","    return topic_details_df"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"uc8xlGJLFeTZ","colab_type":"text"},"source":["You'll be given the function `get_topic_details()` which takes the arguments ldamodel and corpus. It retrieves the details of the topics for each line of text. With that function, you can append the results back to your original data."]},{"cell_type":"code","metadata":{"id":"FGqGREXGXE4J","colab_type":"code","outputId":"acdbe6f5-9cf1-46b8-d26f-0735c6b79791","executionInfo":{"status":"ok","timestamp":1573284849500,"user_tz":-480,"elapsed":728,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":119}},"source":["# Run get_topic_details function and check the results\n","print(get_topic_details(ldamodel, corpus))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["   Dominant_Topic  % Score                                     Topic_Keywords\n","0             3.0   0.9992  investools, stock, go, company, say, year, see...\n","1             4.0   0.9990  enron, mg, plc, rudolph, jane, lon, allen, wol...\n","2             1.0   0.9944  enron, employee, company, million, stock, fund...\n","3             1.0   0.9935  enron, employee, company, million, stock, fund...\n","4             1.0   0.9934  enron, employee, company, million, stock, fund...\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"9ZPF1m4_XE1k","colab_type":"code","outputId":"e266ca8f-8ff4-49f7-d48a-c1a0ed27fafa","executionInfo":{"status":"ok","timestamp":1573284893336,"user_tz":-480,"elapsed":751,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":153}},"source":["# Add original text to topic details in a dataframe\n","contents = pd.DataFrame({'Original text': text_clean})\n","topic_details = pd.concat([get_topic_details(ldamodel, corpus), contents], axis=1)\n","print(topic_details.head())"],"execution_count":0,"outputs":[{"output_type":"stream","text":["   Dominant_Topic  ...                                      Original text\n","0             3.0  ...  [investools, advisory, free, digest, trusted, ...\n","1             4.0  ...  [forwarded, richard, b, sander, hou, pm, justi...\n","2             1.0  ...  [hey, wearing, target, purple, shirt, today, m...\n","3             1.0  ...  [leslie, milosevich, santa, clara, avenue, ala...\n","4             1.0  ...  [rini, twait, e, th, ave, longmont, co, rtwait...\n","\n","[5 rows x 4 columns]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"sJtSSsuEXTAC","colab_type":"code","outputId":"503708d5-2906-4459-ea1c-854ba20f63fd","executionInfo":{"status":"ok","timestamp":1573284901961,"user_tz":-480,"elapsed":757,"user":{"displayName":"Zhenna LU","photoUrl":"","userId":"13596549313510787594"}},"colab":{"base_uri":"https://localhost:8080/","height":153}},"source":["# Create flag for text highest associated with topic 3\n","topic_details['flag'] = np.where((topic_details['Dominant_Topic'] == 3.0), 1, 0)\n","print(topic_details.head())"],"execution_count":0,"outputs":[{"output_type":"stream","text":["   Dominant_Topic  ...  flag\n","0             3.0  ...     1\n","1             4.0  ...     0\n","2             1.0  ...     0\n","3             1.0  ...     0\n","4             1.0  ...     0\n","\n","[5 rows x 5 columns]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"Mw7LNX3yXS9e","colab_type":"code","colab":{}},"source":[""],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"85j0cXhRmOQD","colab_type":"code","colab":{}},"source":[""],"execution_count":0,"outputs":[]}]}