April 2022 • Machine Learning
End-to-end machine learning web application for predicting Bangalore housing prices based on location and amenities. Addresses the complexity of real estate valuation in Bangalore's market where prices vary significantly by location, providing data-driven predictions to help buyers and sellers make informed property decisions.
ML: scikit-learn (Linear Regression, Lasso, Decision Tree), NumPy, Pandas
Backend: Flask, Flask-CORS
Frontend: HTML5, CSS3, JavaScript, Bootstrap 4, jQuery
Deployment: Gunicorn, Nginx
Collected and preprocessed over 13,000 property listings with IQR-based outlier detection and fuzzy matching for Bangalore's complex locality names. Implemented multiple regression algorithms with GridSearchCV hyperparameter tuning, achieving R² score of 0.86 on test data. Built responsive web interface with Flask REST API for model serving, featuring real-time predictions, interactive location selection with autocomplete, and price trend visualizations with average response time under 200ms.
Successfully deployed production model with 86% accuracy and sub-200ms prediction latency. Demonstrated full-stack ML implementation from data collection through deployment, providing accessible real estate valuation tool for data-driven property decisions in Bangalore market.