🚢 Titanic Survival Prediction Using Machine Learning

📌 Overview

This project predicts whether a passenger survived the Titanic disaster using demographic and travel-related information. Multiple machine learning classification models are trained, cross-validated, and compared to select the best-performing model.

The project follows an end-to-end machine learning workflow including data preprocessing, model evaluation, visualization, and final prediction on unseen test data.

---

🎯 Objective

• Predict passenger survival (Survived)
• Perform exploratory data analysis (EDA)
• Train and compare multiple classification models
• Select the best model using F1-score
• Generate predictions for unseen test data

---

📊 Dataset

Source: Kaggle Titanic Dataset

Files Used:

• train.csv – Training & validation data
• test.csv – Unseen data for final prediction

Key Features:

• Pclass – Passenger class
• Sex – Gender
• Age – Age of passenger
• SibSp – Number of siblings/spouses aboard
• Parch – Number of parents/children aboard
• Fare – Ticket fare
• Embarked – Port of embarkation

Target Variable:

• Survived
  • 0 = Did not survive
  • 1 = Survived

---

🧠 Models Used

• Logistic Regression (Baseline)
• Random Forest Classifier
• Support Vector Machine (SVM)
• K-Nearest Neighbors (KNN)
• XGBoost Classifier

---

🔬 Methodology

1. Data loading and inspection
2. Missing value handling and data cleaning
3. Encoding categorical variables
4. Feature scaling
5. Train–validation split
6. 5-fold cross-validation
7. Model comparison using F1-score
8. Best model selection
9. Prediction on test dataset

---

📈 Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1-score (primary metric)
• Confusion Matrix

F1-score was prioritized due to class imbalance in the dataset.

---

📓 Notebook

Complete analysis and implementation is available in the notebook:

---

🚀 How to Run

1️⃣ Install dependencies

pip install -r requirements.txt