🔥 AI-Powered Crime Heatmap An interactive web application that visualizes the predicted danger level of areas across map using a machine learning model trained on historical crime data. Built with React, Flask, and scikit-learn, the app helps identify high-risk areas at different times of day, aiding awareness and prevention.

🚀 Features 📍 Heatmap Visualization: See real-time crime danger levels (red = high risk, yellow = moderate, green = low).

🧠 ML Model Prediction: Random Forest Classifier predicts danger level based on location, hour, and day of week.

🕒 Time-Aware: Query predictions by custom date & time.

📡 REST API: /predict for single-point queries and /heatmap for bounding-box predictions.

🌐 Full-Stack Setup: React frontend + Flask backend integrated with CORS and concurrently.

🛠️ Tech Stack Layer Tools Used Frontend React, Tailwind CSS, Leaflet (Map) Backend Python, Flask, Flask-CORS ML Model scikit-learn (RandomForest) Data Custom CSV-based synthetic crime data Dev Tools concurrently, npm, CORS

📦 Folder Structure bash Copy Edit Crime-Heatmap/ ├── frontend/ # React App (map + UI) ├── backend/ # Flask API (ML model + endpoints) │ └── app.py # Main backend logic │ └── crime_data.csv # Sample training data ├── package.json # Dev script to run both client + server ⚙️ Setup Instructions 🔧 1. Clone the Repository bash Copy Edit git clone https://github.com/your-username/crime-heatmap.git cd crime-heatmap 🖥️ 2. Install Dependencies bash Copy Edit npm install # For root (concurrently) cd frontend && npm install # React frontend cd ../backend && pip install -r requirements.txt # Flask backend 🚀 3. Start the App bash Copy Edit npm run start This runs:

React app on http://localhost:3000

Flask API on http://localhost:5000

🔍 API Endpoints POST /predict Predicts danger level at a single point.

json Copy Edit { "lat": 51.5074, "lng": -0.1278, "hour": 22, "day_of_week": 5 } GET /heatmap Returns all crimes (with predicted danger) within a bounding box.

sql Copy Edit /heatmap?north=...&south=...&east=...&west=...&hour=...&day=... 🧠 AI Model Details Model: RandomForestClassifier

Features: latitude, longitude, hour, day_of_week

Output: danger_level → 0 (Low), 1 (Medium), 2 (High)

Trained on demo data covering Greater London

📊 Demo Dataset Includes a CSV file with synthetic crime data (200 rows) for London, with varied:

Coordinates

Times & dates

Crime types like: Murder, Theft, Littering, etc.

📌 To Do / Ideas Integrate with live crime datasets (e.g. UK Police API)

Add borough-specific clustering

Deploy via Docker or Vercel + Render

Push notifications for high-risk zones

📸 Screenshots (Add screenshots of your React map with red/yellow/green points)

🧑‍💻 Author Made by [Your Name] — feel free to connect or fork!