Aashiyana — House Price Prediction (Delhi NCR)

An end-to-end machine learning pipeline that predicts house prices in Delhi NCR, wrapped in a Flask web app with price prediction, EMI calculator, loan eligibility, and budget planning tools.

Table of Contents

• Project Overview
• Screenshots
• Dataset
• Model Performance
• Project Structure
• Quick Start
• Web App Features
• Deployment

Project Overview

This project:

• Loads and cleans a raw dataset of 39,000+ Delhi NCR property listings
• Engineers features (total rooms, bed/bath ratio, parking/lift/balcony flags)
• Trains and compares Linear Regression, Random Forest, and XGBoost regressors
• Selects the best model by test R² and saves it for inference
• Serves predictions through a Flask web app with a full front-end (search, filters, map view, calculators, auth)

Screenshots

— see Quick Start to run it yourself.

Homepage	Sign Up
Dashboard	AI Price Prediction
Predicted Price Output	EMI Calculator
Loan Eligibility Checker	Budget Planner
Interactive Price Heatmap	Area Unit Converter

Dataset

• Raw data: 39,385 listings (Delhi_v2.csv)
• After cleaning (missing values, outlier removal): 7,435 listings used for training/testing
  • Train: 5,948 rows
  • Test: 1,487 rows

Model Performance (verified, actual run output)

Model	Test RMSE	Test R²
Random Forest (selected model)	₹242,368	0.9978
XGBoost	₹339,024	0.9957
Linear Regression	₹1,116,933	0.9531

Random Forest was selected and saved as house_price_model.pkl.

Note on R²: The R² here is very high because Price_sqft is included as an input feature, and since price ≈ area × Price_sqft, the model is partly reconstructing price from a feature closely tied to price itself. This is expected behavior for this feature set, not an error — worth knowing if asked about it.

Project Structure

├── Delhi_v2.csv              # Raw dataset
├── data_preprocessing.py     # Cleaning + feature engineering
├── model_training.py         # Trains & compares models
├── predict.py                 # Prediction interface
├── utils.py                   # Shared helper functions
├── main.py                    # Runs the full pipeline end-to-end
├── house_price_model.pkl      # Saved best model
├── app.py                     # Flask web app
├── templates/, static/        # Web front-end
├── screenshots/                # App screenshots (see Screenshots section above)
└── DEPLOYMENT.md               # Deployment guide

Quick Start

pip install -r requirements.txt
python main.py          # run the ML pipeline
python app.py            # run the web app

Web App Features

• Property search with filters and map view
• AI-powered price prediction backed by the trained model
• Interactive price heatmap across Delhi NCR
• EMI calculator, loan eligibility checker, budget planner
• Area unit converter (sq.ft, sq.m, bigha, marla, etc.)
• User authentication (login/signup, Google sign-in)

Deployment

See DEPLOYMENT.md for Docker and Railway deployment steps.

License

Provided as-is for educational and portfolio use.

---

Built with Python, scikit-learn, and XGBoost