Abstract:
"This study undertakes the task of developing a comprehensive house price prediction system for Sri Lanka. The system leverages various machine learning regression algorithms to predict house prices based on a range of features such as land size, house size, no. of bedrooms, no. of bathrooms, city, and district. In-depth data exploration, pre-processing, and feature selection precede the application of regression models, ensuring an optimal dataset for analysis.
The model performance has been evaluated using metrics like Mean Absolute Error and R² score, and the best performing models - Gradient Boosting, Random Forest, Extra Trees and K-Nearest Neighbors were integrated into the final system. This system was developed in Django framework, owing to its ease of Python script integration and robustness, offering a user-friendly interface to input the house features and display the average predicted price along with a graphical representation of the prediction results from each model.
Web scraping was employed to automate the data extraction process from online house sale websites, providing an up-to-date dataset for the system. This study contributes to the domain of real estate price prediction by offering a reliable, robust, and user-friendly system to forecast house prices, aiding both sellers and buyers in the market. However, the study acknowledges limitations, including the need for re-training bots for different data sources, negotiation influences on actual prices, and exclusion of environmental variables. Future work is envisioned to overcome these limitations and enhance the model's accuracy and comprehensiveness."
