Titanic Survival Prediction - CS 422 Final Project

Project Overview

This project is part of the CS 422 Course Project , aiming to apply data mining techniques to a real-world predictive modeling problem. The Titanic dataset was used to build a classification model that predicts whether a passenger survived the infamous Titanic disaster based on available features.

Problem Statement

Given the passenger data (such as age, sex, ticket class, and number of siblings/spouses aboard), predict whether a passenger survived (1) or not (0). This is a supervised binary classification problem.

Methodology Summary

1. Data Preprocessing

• Dropped irrelevant columns: PassengerId, Name, Ticket
• Imputed missing values:
  • Numeric: median
  • Categorical: most frequent
• Standardized numerical features
• One-hot encoded categorical variables
• Engineered new features:
  • FamilySize = SibSp + Parch + 1
  • IsAlone = 1 if FamilySize == 1 else 0

2. Modeling

• Primary model: Random Forest Classifier
• Baseline models: Logistic Regression, Support Vector Machine (SVM)
• Hyperparameter tuning via GridSearchCV (5-fold cross-validation)
• Evaluation metrics:
  • Accuracy
  • ROC-AUC
  • Confusion Matrix
  • Feature Importance
  • ROC Curve

Key Results

Model	Accuracy	ROC-AUC
Random Forest	0.8045	0.8431
Logistic Regression	0.8045	0.8267
SVM	0.8212	0.8372

• Random Forest performed best overall and was chosen as the final model.
• Key features: Sex, Fare, Age, Pclass, IsAlone, Embarked_C