AI-Based Restaurant Location Intelligence System

Live Demo

Open Live App

Production-style Data Science project for restaurant businesses planning to open in Chennai, India. Uses machine learning, clustering, geospatial analysis, and interactive business intelligence to recommend optimal locations.

---

Overview

This system analyzes 11,800+ Zomato restaurants across Chennai to help entrepreneurs answer:

• Where should I open my restaurant?
• What rating can I expect?
• How intense is the competition?
• Which cuisines perform best in each area?

Key Capabilities

Module	Technique	Purpose
Rating Prediction	CatBoost Regressor	Predict expected dining rating
Area Segmentation	KMeans Clustering	Market segment classification
Competition Analysis	DBSCAN	Geospatial competition hotspots
Recommendation Engine	Composite Scoring	Rank best areas for new outlets

---

Project Structure

├── data/                          # Raw & processed datasets
│   └── Zomato_Chennai_Final.csv
├── notebooks/                     # Jupyter notebooks (optional EDA)
├── models/                        # Trained ML artifacts
├── src/
│   ├── config.py                  # Paths & constants
│   ├── preprocessing/             # Data cleaning
│   ├── features/                  # Feature engineering
│   ├── modeling/                  # CatBoost rating model
│   ├── clustering/                # KMeans & DBSCAN
│   ├── recommendation/            # Location recommendation engine
│   ├── visualization/             # Plotly & Folium charts
│   └── utils/                     # Helper functions
├── app/
│   ├── streamlit_app.py           # Main dashboard
│   └── components/                # UI & data loaders
├── outputs/
│   ├── figures/                   # Saved charts
│   ├── maps/                      # HTML maps
│   └── reports/                   # CSV & JSON reports
├── main.py                        # Training pipeline
├── requirements.txt
└── README.md

---

Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Train Models

python main.py

This runs:

1. Data cleaning
2. Feature engineering (area-level business metrics)
3. KMeans area clustering
4. DBSCAN competition analysis
5. CatBoost rating model training

3. Launch Dashboard

streamlit run app/streamlit_app.py

Open http://localhost:8501 in your browser.

---

Dashboard Features

Input (Sidebar)

• Cuisine type
• Restaurant category (Restaurant, Cafe, Fast Food, etc.)
• Minimum expected rating threshold

Output

• Best areas to open a restaurant (ranked)
• Predicted success score (0–100)
• Expected rating (CatBoost)
• Competition level (Low / Medium / High)
• Area demand score
• Nearby similar restaurants
• Interactive Plotly charts
• Folium geospatial maps
• Downloadable Excel reports

---

Machine Learning Pipeline

Feature Engineering

Feature	Description
restaurant_count_per_area	Total restaurants in area
avg_area_rating	Mean dining rating
cuisine_popularity_score	City-wide cuisine demand
competition_density	Competition intensity proxy
premium_restaurant_ratio	Share of rating ≥ 4.2
area_success_score	Composite business score

Models

1. CatBoost Regressor — Predicts dining_rating using area, cuisine, category, and engineered features
2. KMeans (k=6) — Segments areas into market clusters
3. DBSCAN — Identifies geospatial competition hotspots vs. opportunity zones

Recommendation Score

Score = 25% Demand + 20% Quality + 20% Cuisine Fit
      + 20% Low Competition + 15% Geo Opportunity

---

Dataset

Source: Zomato_Chennai_Final.csv

Column	Description
name_of_restaurant	Restaurant name
market_segment	Category (Restaurant, Cafe, etc.)
cuisine	Comma-separated cuisines
area/location	Chennai neighborhood
latitude, longitude	Geo coordinates
dining_rating	Zomato dining rating (0–5)

---

Tech Stack

• Frontend: Streamlit
• Backend: Python
• ML: scikit-learn, CatBoost, XGBoost
• Viz: Plotly, Matplotlib, Seaborn, Folium
• Geo: geopy, Folium

---

Use Cases

• Final year engineering project
• GitHub portfolio showcase
• Data Science internship interviews
• Restaurant business feasibility studies

---

Author Notes

Built as a modular, scalable data science product with separated concerns:

• preprocessing → features → modeling → clustering → recommendation → visualization → app

Run python main.py after any data updates to refresh models.

---

License

MIT — Open for educational and portfolio use.