Titanic Survival Predictor

A full-stack machine learning web app that predicts Titanic passenger survival using 8 classifiers simultaneously, built from the original exploratory notebook.

---

What's New

The original ML-Titanic.ipynb notebook has been rebuilt as a modern full-stack web application:

Before	After
Jupyter Notebook	FastAPI backend + React frontend
Loads local CSV files	Bundled dataset, served via REST API
Static cell output	Live interactive UI with real-time predictions
Manual model comparison	Model leaderboard with accuracy bars
Matplotlib/Seaborn charts	Interactive Recharts bar charts

---

Features

• Survival Predictor — enter passenger details (class, sex, age, fare, port) and get an instant prediction
• 8-model consensus — all classifiers vote; result shown with probability bar and per-model breakdown
• Model Leaderboard — ranked accuracy table for all 8 classifiers
• EDA Dashboard — survival rates by sex, passenger class, age group, and title (interactive charts)

---

Tech Stack

Backend

• Python 3.9+
• FastAPI — REST API
• scikit-learn — ML models
• pandas + NumPy — data processing

Frontend

• React 18 + TypeScript
• Vite — build tool
• Tailwind CSS — styling
• Recharts — charts

---

ML Models & Accuracy

Rank	Model	Accuracy
🥇	Random Forest	93.6%
🥈	Decision Tree	93.6%
🥉	K-Nearest Neighbors	87.1%
4	Support Vector Machine	83.5%
5	Logistic Regression	80.0%
6	Linear SVC	79.7%
7	Naive Bayes	77.0%
8	Perceptron	66.6%

---

Feature Engineering

Replicates the original notebook's pipeline:

• Title extraction from passenger names (Mr / Mrs / Miss / Master / Rare)
• Age imputation via median by Sex × Pclass grid
• Age banding into 5 bins (0–16, 16–32, 32–48, 48–64, 64+)
• Categorical encoding for Sex and Embarked
• Models trained on: Pclass, Sex, Age, Fare, Embarked, Title

---

Project Structure

├── ML-Titanic.ipynb          # Original exploratory notebook
├── backend/
│   ├── main.py               # FastAPI app (3 endpoints)
│   ├── pipeline.py           # Feature engineering + model training
│   ├── data/
│   │   └── titanic.csv       # Training dataset (891 rows)
│   └── requirements.txt
└── frontend/
    ├── src/
    │   ├── App.tsx            # Tab layout (Predict / Models / Stats)
    │   ├── components/
    │   │   ├── PredictionForm.tsx
    │   │   ├── ResultCard.tsx
    │   │   ├── ModelLeaderboard.tsx
    │   │   └── StatsDashboard.tsx
    │   └── types.ts
    ├── vite.config.ts         # Proxies /api → localhost:8000
    └── package.json

---

Setup & Run

1. Backend

cd backend
pip install -r requirements.txt
python3 -m uvicorn main:app --port 8000

Models train automatically on startup (~2 seconds). API available at http://localhost:8000.

2. Frontend

cd frontend
npm install
npm run dev

Open http://localhost:5173

---

API Endpoints

Method	Endpoint	Description
POST	/api/predict	Predict survival for a passenger
GET	/api/models	Accuracy scores for all 8 models
GET	/api/stats	EDA stats (survival rates by group)

Predict request body:

{
  "pclass": 1,
  "sex": "female",
  "age": 30,
  "sibsp": 0,
  "parch": 0,
  "fare": 100,
  "embarked": "S"
}

---

Dataset

The Titanic dataset — 891 passengers, 38.4% survival rate. Features: PassengerId, Survived, Pclass, Name, Sex, Age, SibSp, Parch, Ticket, Fare, Cabin, Embarked.