The RealEstiMate system is a comprehensive application developed as part of my BSc (Hons) Computer Science final project at the University of Plymouth. It combines machine learning for property price predictions with a full-featured web application to address inefficiencies in manual real estate valuations, enabling informed decisions for buyers, sellers, and investors. Data is fetched in real-time from a database for dynamic analysis and visualizations.

Key Features:

• Utilized multiple regression models (Linear Regression, Random Forest, XGBoost, Gradient Boosting, LightGBM, CatBoost) to test for the best results, achieving high accuracy with LightGBM, which was selected as the final model to predict property prices based on features like location, size, year built, and economic factors (e.g., inflation rate).
• Data preprocessing, model training, and evaluation using Melbourne housing dataset, including handling missing values, label encoding, and dataset split (80/20).
• Model optimization with Bayesian Optimization for hyperparameters and SHAP for feature importance visualization.
• Full web application with user authentication (login/signup/signout) and role-based access.
• Price Calculator page for inputting property details and receiving instant predictions.
• Interactive Maps for visualizing property locations and market insights.
• Real-time Trends and Patterns page showing price changes over time with line graphs (using Recharts) and data fetched from the database.
• Categorical Feature Importance Analysis to understand key drivers of property prices.
• Supports real-time data storage and retrieval, seamless integration between frontend and backend.
• Evaluation metrics: RMSE, MAE, R² Score to ensure accuracy and reliability.

Technology Stack:

• Frontend: React.js, HTML, CSS, JavaScript, Recharts (for visualizations)
• Backend: Node.js (with Express), Flask (for ML integration)
• Database: Firebase (real-time)
• Programming Language: Python (for ML), JavaScript (for web)
• Libraries: Pandas, NumPy, Seaborn, Matplotlib, Scikit-learn, LightGBM, SHAP, Bayesian Optimization, Axios (for API calls)
• Tools: Jupyter Notebook for ML development, Postman for API testing
• Other: Data visualization (e.g., SHAP summary plots, line charts), model saving/loading with Pickle

This project demonstrates my skills in full-stack web development, machine learning, data analysis, predictive modeling, real-time data handling, and end-to-end system design from data preparation to deployment, contributing to advancements in real estate technology.