Real Estate Price Prediction in Prague

The bachelor’s thesis focuses on the prediction of market prices of residential properties in the city of Prague. The aim was to design and evaluate a predictive model that estimates pro- perty value based on technical characteristics, location attributes, and accessibility of public infrastructure. Data were obtained via web scraping from a public interface of a real estate platform (referred to hereinafter as "RealityXYZ"), with prior consultation and approval from the company. The methodology is based on the KDD (Knowledge Discovery in Databases) framework and covers the entire process from data selection and preprocessing through transformation to the training and tuning of regression models. Various machine learning algorithms were compared, with the best performance achieved by ensemble methods—specifically, Random Forest and Gradient Boosting. To improve the quality of location-based inputs, geolocation data from external APIs were integrated. The results show that incorporating features such as distance to the nearest metro station, city district, and administrative borough significantly improves model accuracy. The proposed approach achieved a relative mean absolute error (MAE) below 14% for apartment predictions and can serve as a foundation for a practical tool for urban price estimation.