title	author	output
Machine Learning for Binary Classification	Tai Luan Nguyen	pdf_document html_document default default

Overview

This project performs various machine learning tasks on a dataset with 19020 samples and 11 attributes. Demonstrates the application of various machine learning algorithms to a dataset, allowing the comparison of their performance using classification reports.

Additionally, it conducts a hyperparameter search for a neural network model.

The project including:

• Importing Libaries
• Dataset Loading and Preprocessing
• Split dataset
• Data Scaling and Oversampling
• k-Nearest Neighbour (kNN) Classifier
• Naive Bayes Classifier
• Logistic Regression Classifier
• Support Vector Machine (SVM) Classifier
• Neurol Network (Deep Learning) Classifier

Project link: https://github.com/luan30092000/binaryClassification

Dataset

Data are MC generated to simulate registration of high energy gamma particles in an atmospheric Cherenkov telescope.

Reference: Bock,R.. (2007). MAGIC Gamma Telescope. UCI Machine Learning Repository. https://doi.org/10.24432/C52C8B.

Importing Libaries

Necessary libraries and its purpose for this project:\

• numpy: data manipulation and array operation.
• pandas: load and manipulate dataset, as well as for data preprocessing and analysis.
• matplotlib: matplotlib.pyplot, create histograms for data visualization.
• sklearn.preprocessing: StandardScaler from this library is used for feature scaling, which standardizes the data to have a mean of 0 and a standard deviation of 1.
• imblearn.over_sampling: RandomOverSampler is used to oversample the minority class to balance the class distribution.
• sklearn.neighbors: KNeighborsClassifier to create and evaluate a kNN model.
• sklearn.naive_bayes: GaussianNB to create and evaluate a Naive Bayes model.
• sklearn.linear_model: LogisticRegression to create and evaluate a logistic regression model.
• sklearn.svm: SVC to create and evaluate an SVM model
• tensorflow: used to create and train a neural network for classification

Dataset Loading and Preprocessing

• Loads a dataset from a file named "magic04.data" using pandas and defines column names for the dataset.
• Converts the "class" column to binary values (0 or 1) by mapping "g" to 1 and "h" to 0.
• Plots histograms for each feature, comparing the distributions for class 0 (hadron) and class 1 (gamma).

Train, Validation, and Test Datasets:

• The dataset is split into train, validation, and test sets using the np.split function, with a 60-20-20 split ratio.

Data Scaling and Oversampling

• Scale the features and, optionally, oversample the dataset.
• StandardScaler is used to scale the features

k-Nearest Neighbours (kNN) Classifier

• Uses the scikit-learn library to create a k-Nearest Neighbors (kNN) model with 5 neighbors.
• Fits the model to the training data and makes predictions on the test data.
• Accuracy: 81%

Naive Bayes Classifier

• Uses scikit-learn's Gaussian Naive Bayes classifier to create a Naive Bayes model.
• Fits the model to the training data and makes predictions on the test data.
• Accuracy: 73%

Logistic Regression Classifier

• Logistic regression model is created using scikit-learn
• The model is trained on the training data and used to predict on the test data.
• Accuracy: 78%

Support Vector Machine (SVM) Classifier

• Support Vector Machine (SVM) classifier is created using scikit-learn.
• Model is trained on the training data and used to predict on the test data.
• Accuracy: 85%

Neural Network (Deep Learning) Classifier

• Neural network model is defined using TensorFlow/Keras
• Performs a hyperparameter grid search, iterating over various combinations of the number of nodes, dropout probability, learning rate, and batch size.
• Trains multiple neural network models with different hyperparameters on the training data and selects the model with the lowest validation loss.
• The chosen neural network model has the accuracy of 87%