notebooks/spam_detector.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# Spam detector\n",
        "\n",
        "**Predicting the probability that a given email is a spam email**\n",
        "\n",
        "**Natural Language Processing. Supervised Learning. Binary classification**\n",
        "\n",
        "Data from [Applied Text Mining in Python | Coursera](https://www.coursera.org/learn/python-text-mining/)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Path changed to /home/angel/DRIVE/data-science-keras\n",
            "\n",
            "HARDWARE:\n",
            "CPU:\tIntel(R) Core(TM) i7-9750H CPU @ 2.60GHz\n",
            "RAM:\t8 GB\n",
            "GPU:\tNVIDIA GeForce RTX 2070\n",
            "\n",
            "OS:\tLinux-5.10.16.3-microsoft-standard-WSL2-x86_64-with-glibc2.35\n",
            "\n",
            "ENV:\t/home/angel/miniconda3/envs/gpu\n",
            "\n",
            "PYTHON:\t3.10.4 (main, Mar 31 2022, 08:41:55) [GCC 7.5.0]\n",
            "tensorflow                    2.9.1\n",
            "pandas                        1.4.3\n",
            "numpy                        1.23.1\n",
            "\n",
            "EXECUTION PATH: /home/angel/DRIVE/data-science-keras\n",
            "EXECUTION DATE: Mon Aug 15 13:22:49 2022\n"
          ]
        }
      ],
      "source": [
        "%matplotlib inline\n",
        "import sys\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "sys.path.append(\"../\")\n",
        "import ds_boost\n",
        "\n",
        "ds_boost.set_parent_execution_path()\n",
        "ds_boost.info_system()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 1. Data Processing"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "5572 rows \n",
            "2 columns\n"
          ]
        }
      ],
      "source": [
        "df = pd.read_csv(\"data/spam.csv\")\n",
        "print(\"{} rows \\n{} columns\".format(*df.shape))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "#### Explore and clean the target"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Ratio of email spam: 0.134\n"
          ]
        },
        {
          "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>text</th>\n",
              "      <th>target</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Go until jurong point, crazy.. Available only ...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Ok lar... Joking wif u oni...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Free entry in 2 a wkly comp to win FA Cup fina...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                text  target\n",
              "0  Go until jurong point, crazy.. Available only ...       0\n",
              "1                      Ok lar... Joking wif u oni...       0\n",
              "2  Free entry in 2 a wkly comp to win FA Cup fina...       1"
            ]
          },
          "execution_count": 3,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# print(df['target'].squeeze().value_counts(dropna=False))\n",
        "df[\"target\"] = np.where(df[\"target\"] == \"spam\", 1, 0)  # change target ('ham', 'spam') to (0,1)\n",
        "print(\"Ratio of email spam: {:.3f}\".format(np.mean(df[\"target\"])))\n",
        "df.head(3)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "####  Split data into training and test set"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {},
      "outputs": [],
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "df_train, df_test = train_test_split(df, test_size=0.3, stratify=df[\"target\"], random_state=0)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "#### Transform data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "train size \t X:(3900, 10) \t Y:(3900, 2)\n",
            "test size  \t X:(1672, 10) \t Y:(1672, 2) \n"
          ]
        }
      ],
      "source": [
        "import keras\n",
        "from keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "\n",
        "max_words = 1000\n",
        "max_length = 10\n",
        "\n",
        "x_train, y_train = df_train[\"text\"], df_train[\"target\"]\n",
        "x_test, y_test = df_test[\"text\"], df_test[\"target\"]\n",
        "\n",
        "tokenizer = Tokenizer(num_words=max_words)\n",
        "# non-alphanumeric characters could be important here (filters=''), but better results were\n",
        "#  obtained with default filters\n",
        "\n",
        "tokenizer.fit_on_texts(x_train)\n",
        "\n",
        "x_train = tokenizer.texts_to_sequences(x_train)\n",
        "x_test = tokenizer.texts_to_sequences(x_test)\n",
        "\n",
        "# padding\n",
        "x_train = pad_sequences(x_train, max_length)\n",
        "x_test = pad_sequences(x_test, max_length)\n",
        "\n",
        "# one-hot encoding the target\n",
        "num_classes = 2\n",
        "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
        "y_test = keras.utils.to_categorical(y_test, num_classes)\n",
        "\n",
        "print(\"train size \\t X:{} \\t Y:{}\".format(x_train.shape, y_train.shape))\n",
        "print(\"test size  \\t X:{} \\t Y:{} \".format(x_test.shape, y_test.shape))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 2. Neural Network"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding (Embedding)       (None, 10, 10)            10000     \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 100)               0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 10)                1010      \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 10)                0         \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 2)                 22        \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 11,032\n",
            "Trainable params: 11,032\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Training ....\n",
            "Epoch 20: early stopping\n",
            "CPU times: user 8.21 s, sys: 3.67 s, total: 11.9 s\n",
            "Wall time: 8 s\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1008x288 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "\n",
            "Training loss:  \t0.0333\n",
            "Validation loss: \t 0.0596\n",
            "\n",
            "Training accuracy: \t0.993\n",
            "Validation accuracy:\t0.984\n"
          ]
        }
      ],
      "source": [
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten\n",
        "from keras.layers import Embedding\n",
        "\n",
        "model = Sequential()\n",
        "model.add(Embedding(max_words, 10, input_length=max_length))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(10, activation='relu'))\n",
        "model.add(Dropout(0.25))\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()\n",
        "\n",
        "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "\n",
        "print('Training ....')\n",
        "callbacks = [keras.callbacks.EarlyStopping(monitor='val_loss', patience=1, verbose=1)]\n",
        "%time history = model.fit(x_train, y_train, batch_size=128, epochs=30, verbose=0, \\\n",
        "                          validation_split = 0.3, callbacks=callbacks)\n",
        "\n",
        "ds_boost.show_training(history)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "#### Evaluate Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "\n",
            "Test Accuracy: 0.980\n",
            "\n",
            "53/53 [==============================] - 0s 3ms/step\n",
            "Neural Network ROC AUC: 0.988 \n",
            "\n"
          ]
        }
      ],
      "source": [
        "score = model.evaluate(x_test, y_test, verbose=0)\n",
        "print(\"\\nTest Accuracy: {:.3f}\\n\".format(score[1]))\n",
        "\n",
        "y_pred = model.predict(x_test)\n",
        "\n",
        "from sklearn.metrics import roc_auc_score\n",
        "\n",
        "print(\"Neural Network ROC AUC: {:.3f} \\n\".format(roc_auc_score(y_test, y_pred)))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 3. Compare with classical vectorizer + Multinomial Naive Bayes"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "\n",
            "MultinomialNB ROC_AUC: 0.958\n"
          ]
        }
      ],
      "source": [
        "from sklearn.feature_extraction.text import CountVectorizer\n",
        "from sklearn.naive_bayes import MultinomialNB\n",
        "\n",
        "x_train, y_train = df_train[\"text\"], df_train[\"target\"]\n",
        "x_test, y_test = df_test[\"text\"], df_test[\"target\"]\n",
        "\n",
        "vect = CountVectorizer().fit(x_train)\n",
        "\n",
        "x_train_vectorized = vect.transform(x_train)\n",
        "\n",
        "model = MultinomialNB(alpha=0.2).fit(x_train_vectorized, y_train)\n",
        "\n",
        "predictions = model.predict(vect.transform(x_test))\n",
        "\n",
        "print(\"\\nMultinomialNB ROC_AUC: {:.3f}\".format(roc_auc_score(y_test, predictions)))"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3.10.4 ('gpu')",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.10.4"
    },
    "vscode": {
      "interpreter": {
        "hash": "d2efdb989dc164e4378df2c802d4acdf5890bc9f1d40e64af3f699a110ebaaa8"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}