XavierNumbus/real-estate-price-prediction

Real Estate Price Prediction

A multi-page Streamlit web application for Gurgaon real estate market analysis, price prediction, and apartment recommendation.

Features

Tech Stack

• —Frontend: Streamlit
• —ML / Data: scikit-learn, pandas, NumPy, pickle
• —Visualization: Plotly Express, Matplotlib, Seaborn, WordCloud
• —Dataset: Gurgaon real estate listings with sector-level geo data

Project Structure

CapstoneProject/
├── Home.py                          # Landing page
├── pages/
│   ├── Price Predictor.py           # ML-based price prediction
│   ├── Analysis App.py              # Data analytics & visualizations
│   └── 3_Recommend Appartments.py   # Location search & apartment recommender
├── datasets/
│   ├── data_viz1.csv                # Cleaned dataset for visualizations
│   ├── feature_text.pkl             # Feature text for word cloud
│   ├── location_distance.pkl        # Inter-property distance matrix
│   ├── cosine_sim1.pkl              # Cosine similarity matrix (amenities)
│   ├── cosine_sim2.pkl              # Cosine similarity matrix (features)
│   └── cosine_sim3.pkl              # Cosine similarity matrix (price/location)
├── df.pkl                           # Processed dataframe for dropdowns
├── pipeline.pkl                     # Trained ML pipeline (stored via Git LFS)
└── README.md

How It Works

Price Predictor

Takes user inputs — property type, sector, BHK, bathrooms, balconies, age, area, furnishing, luxury category, and floor — and feeds them into a serialized scikit-learn pipeline to predict price with a ±0.22 Cr confidence band.

Analytics Dashboard

• —Geo Map: Sector-wise average price per sqft plotted on an interactive map
• —Word Cloud: Most common amenities/features across listings
• —Area vs Price: Scatter plot colored by BHK count
• —BHK Distribution: Sector-level pie chart
• —Price Range: Box plot comparing price ranges across BHK types
• —Property Type Distribution: KDE histogram comparing flats vs houses

Apartment Recommender

• —Radius Search: Lists all properties within a user-specified radius (km) from a selected location
• —Similar Properties: Uses a weighted combination of 3 cosine similarity matrices to recommend the top 5 most similar apartments

Installation & Setup

# 1. Clone the repository
git clone https://github.com/Xavier-Numbus/Real-Estate-price-prediction-.git
cd Real-Estate-price-prediction-

# 2. Install Git LFS (required for pipeline.pkl)
git lfs install
git lfs pull

# 3. Create a virtual environment
python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate

# 4. Install dependencies
pip install streamlit pandas numpy scikit-learn plotly wordcloud matplotlib seaborn

# 5. Run the app
streamlit run Home.py

Usage

1. 1.Open the app in your browser (default: http://localhost:8501)
2. 2.Use the sidebar to navigate between pages
3. 3.On Price Predictor — fill in property details and click Predict
4. 4.On Analytics — explore interactive charts and maps
5. 5.On Recommend Apartments — enter a location and radius to search, or select an apartment for similar recommendations

Dataset

The dataset covers residential properties (flats and houses) across multiple sectors in Gurgaon, Haryana, India. Features include:

• —Property type, sector, BHK, bathrooms, balconies
• —Built-up area (sq ft), property age, servant/store rooms
• —Furnishing type, luxury category, floor category
• —Latitude/longitude for geo-visualization
• —Price (in Crores INR)