Genomic Analysis & Predictive Modeling Framework

This repository contains a collection of Jupyter Notebooks focused on applying advanced Machine Learning (ML) techniques to genomic datasets. The primary objective is to classify genetic disorders and subclasses using state-of-the-art boosting algorithms and unsupervised learning methods.

🚀 Project Overview

The core of this project revolves around the Genetic Disorder Dataset, where multiple models were trained and evaluated to achieve high-precision classification. The workflow includes data preprocessing, feature engineering, hyperparameter tuning, and comparative model analysis.

🛠 Tech Stack

• Languages: Python
• ML Frameworks: Scikit-Learn, XGBoost, CatBoost, LightGBM
• Data Analysis: Pandas, NumPy
• Visualization: Matplotlib, Seaborn

📂 Codebase Breakdown

🧬 Genetic Disorder Classification (Supervised)

• Model Training v3 & v4: Implementation of LightGBM and XGBoost with GPU acceleration support for multi-output classification.
• CatBoost Models: High-performance gradient boosting optimized for categorical genomic features, including CPU and GPU training variants.
• Random Forest (RFC): A robust ensemble baseline for disorder classification using scikit-learn.

🔍 Discovery & Clustering (Unsupervised)

• KMeans on Genomes: Unsupervised clustering to identify patterns within encoded genomic data.
• Apriori Algorithm: Association rule mining to discover hidden relationships between specific genes and disorders.
• Geographical KMeans: A separate implementation focused on country-level geographical clustering.

📊 Exploratory & Lab Work

• DS&A Lab05: foundational regression analysis and data exploration techniques.

📈 Performance & Evaluation

Models are evaluated using:

• Stratified K-Fold Cross-Validation
• Confusion Matrices
• Macro-averaged F1-Scores
• ROC/AUC Curves

📋 How to Use

1. Clone the repository.
2. Ensure you have the required libraries installed: pip install xgboost catboost lightgbm scikit-learn pandas matplotlib.
3. The datasets (e.g., train_encoded.csv) are expected in the path specified within the notebooks or can be updated to your local directory.