Titanic Survival Analysis

Titanic Survival Analysis#

🚢 RMS Titanic: A Data Science Case Study

Exploring survival patterns through comprehensive machine learning analysis

Project Overview#

The RMS Titanic disaster offers a tragic but data-rich case study in survival analysis. On April 15, 1912, the “unsinkable” ship struck an iceberg and sank, resulting in the deaths of over 1,500 passengers and crew. The disaster has been extensively documented, providing a comprehensive dataset that allows us to explore patterns in survival.

This project demonstrates a complete machine learning workflow, from initial data exploration through model development to final interpretation. We’ll investigate which factors most strongly predicted survival and compare multiple classification approaches.

Project Structure#

This analysis is organized into four interconnected notebooks:

📊 1. Data Exploration

Explore →

Initial profiling, univariate and bivariate analysis, missing data patterns

🧹 2. Data Cleaning

Explore →

Handling missing values, outlier detection, feature engineering

🤖 3. Modeling

Explore →

Training and comparing multiple classification algorithms

📈 4. Results

Explore →

Model comparison, interpretation, and final insights

Key Questions#

  • 🔍 Which demographic factors were most predictive of survival?

  • 🔍 How did class, gender, and age interact in determining survival outcomes?

  • 🔍 Which machine learning model best captures the underlying patterns?

  • 🔍 What can we learn about the “women and children first” evacuation policy from the data?

Dataset#

The dataset contains information on 891 passengers, including:

👥 Demographics
Age, Sex, Embarked port
💰 Socioeconomic
Passenger class (Pclass), Fare
👨‍👩‍👧‍👦 Family
Siblings/spouses, Parents/children
🎯 Target
Survival (0 = No, 1 = Yes)

Technical Approach#

We’ll employ a rigorous methodology:

📊 Exploratory Data Analysis
Automated profiling and custom visualizations
🔧 Statistical Imputation
Multiple imputation for missing age values
🤖 Model Comparison
Logistic regression, tree-based methods, ensemble techniques
🔍 Model Interpretation
Feature importance, SHAP values, partial dependence plots
✅ Validation
Stratified cross-validation with multiple evaluation metrics

Expected Outcomes#

By the end of this analysis, we’ll have:

✅ Identified the strongest predictors of survival
✅ Trained and compared multiple classification models
✅ Provided interpretable explanations for model predictions
✅ Documented a reproducible analytical workflow
Contents