Titanic Survival Prediction (AI Capstone Project)

Overview

This project presents an end-to-end machine learning pipeline to predict passenger survival on the Titanic dataset. It demonstrates the complete workflow from data preprocessing and exploratory analysis to model building, evaluation, and prediction.

---

Objective

To build a robust classification model that predicts whether a passenger survived the Titanic disaster based on demographic and travel-related features.

---

Dataset

The dataset includes passenger information such as:

• Age
• Gender
• Passenger Class (Pclass)
• Fare
• Embarked Location

Note: Due to file size and repository best practices, datasets are not included. They can be downloaded from the Kaggle Titanic dataset.

---

Workflow

1. Data Preprocessing

• Handled missing values using median and mode imputation
• Removed irrelevant features (Cabin, Name, Ticket)
• Encoded categorical variables (Sex, Embarked)

2. Feature Engineering

• Created new features such as:

  • FamilySize
  • IsAlone

3. Exploratory Data Analysis (EDA)

• Analyzed survival distribution
• Examined relationships between survival and key features such as gender and passenger class

4. Model Development

Trained and compared multiple machine learning models:

• Logistic Regression
• Decision Tree
• Random Forest
• Gradient Boosting

5. Model Evaluation

• Performed train-test split for validation

• Evaluated models using:

  • Accuracy
  • Confusion Matrix
  • Classification Report

6. Hyperparameter Tuning

• Used GridSearchCV to optimize Random Forest parameters
• Improved model performance through cross-validation

---

Results

• Best Model: Random Forest
• Validation Accuracy: ~82%
• Key Predictors: Gender, Passenger Class, Fare

---

Tech Stack

• Python
• Pandas
• NumPy
• Scikit-learn
• Matplotlib
• Seaborn

---

How to Run

1. Install dependencies:

pip install pandas numpy scikit-learn matplotlib seaborn

2. Run the notebook:

Titanic_Survival_Prediction.ipynb

---

Project Structure

• Titanic_Survival_Prediction.ipynb → Main notebook
• submission.csv → Model predictions
• README.md → Project documentation

---

Key Insights

• Survival probability is strongly influenced by gender and passenger class
• Feature engineering improves model performance
• Ensemble models outperform simpler models

---

Conclusion

This project demonstrates a complete machine learning workflow, highlighting the importance of preprocessing, feature engineering, model comparison, and evaluation in building effective predictive models.

---

Future Improvements

• Incorporate advanced feature engineering (Title extraction, family grouping)
• Experiment with boosting algorithms such as XGBoost or LightGBM
• Improve model interpretability

---

Author

Lavanya Dive BSc Data Science Student