Titanic Survival Analysis

A comprehensive machine learning analysis of the Titanic dataset to predict passenger survival using advanced feature engineering and ensemble methods.

Overview

This project implements a complete data science pipeline for the Titanic survival prediction challenge, achieving competitive accuracy through systematic feature engineering and model optimization.

Key Features

• Advanced Feature Engineering: Title extraction, family size indicators, age grouping, fare binning, and interaction features
• Model Optimization: Hyperparameter tuning using GridSearchCV for multiple algorithms
• Ensemble Methods: Voting classifier combining Random Forest, Gradient Boosting, AdaBoost, and SVM
• Performance Analysis: Feature importance analysis and comprehensive model comparison

Models Implemented

• Random Forest Classifier
• Gradient Boosting Classifier
• AdaBoost Classifier
• Support Vector Machine (with scaling pipeline)
• Voting Ensemble

Results

The optimized ensemble model achieves 83.28% accuracy through:

• Enhanced feature engineering (+3.02% improvement)
• Hyperparameter optimization (+1.46% improvement)
• Feature scaling for SVM (+14.14% improvement)
• Ensemble voting methods

Usage

1. Ensure datasets are in the data/ directory
2. Run all cells in Titanic_Survival_Analysis.ipynb
3. Final predictions will be saved to Submission.csv

Author

Yassine Erradouani
yerradouani.me