Machine Learning Projects

This repository contains a collection of my machine learning projects. Each project contains a well documented Jupyter Notebook (.ipynb) and its corresponding HTML export. These projects were accomplished during Data Analysis and Machine Learning graduate course at University of Toronto.

1️⃣ Project 1: Anomaly Detection in Forest Cover Data

Overview

This project focuses on detecting anomalies in forest cover data using statistical modeling techniques. The dataset contains 10 features with a small percentage of anomalies (0.96%).

Key Components

• Dataset: Forest Cover Anomaly Dataset

• Analysis Methods:

  • Single & Multi-feature Gaussian models

  • One & Two-Gaussian approaches

  • Mixture of Gaussians for improved modeling

• Evaluation Metrics: F1-score, AUC

2️⃣ Project 2: Diabetes Severity Classification and Feature Selection

Overview

This project focuses on predicting diabetes severity using a dataset containing various health features. The dataset includes 10 features and the target is based on whether diabetes severity is above or below the median value.

Key Components

• Dataset: Diabetes Health Indicators Dataset

• Analysis Methods:

  • KNN Classification
  • Feature selection using decision trees
  • Cross-validation for hyperparameter tuning

3️⃣ Project 3: Vaccination Data Analysis

Overview

This project focuses on analyzing COVID-19 vaccination data for multiple countries and performing dimensionality reduction techniques to explore trends and reconstruct time-series data. The dataset contains vaccination data for different countries across various dates.

Key Components

• Dataset: COVID-19 Vaccination Data

• Analysis Methods:

  • Time-Series Plotting: Plot time-series for specified countries and standardize the data.
  • Principal Component Analysis (PCA): Compute covariance matrix, eigenvalues, eigenvectors, and visualize the contribution of principal components to explain the dataset's variance.
  • Singular Value Decomposition (SVD): Perform data decomposition using SVD and compare it with PCA results.

4️⃣ Project 4: Superconductor Critical Temperature Prediction

Overview

This project focuses on predicting the critical temperature of a superconductors using a dataset that contains over 20,000 instances, with wide set of features. The goal is to predict the critical temperature (critical_temp) based on these features. The prediction model is implemented using linear regression with different techniques for optimization, including direct solution, full batch gradient descent, and mini-batch gradient descent.

Key Components

• Dataset: Superconductor Dataset.

• Analysis Methods:

  • Linear Regression: Direct solution using matrix inversion.
  • Full Batch Gradient Descent: Implement gradient descent and track training time and RMSE evolution over epochs.
  • Mini-Batch Gradient Descent: Implement mini-batch gradient descent for various batch sizes and investigate convergence.
  • Learning Rate Sensitivity: Investigate the effect of learning rate on convergence for various batch sizes.