HOUSING PRICE PREDICTION USING XGBOOST AND RANDOM FOREST METHODS

Real estate prices often experience significant variations due to various reasons such as shifts in land value and local infrastructure development. An application for predicting housing prices, taking into account the property and neighborhood characteristics and infrastructure, is essential for providing customers with accurate insights into where and when to invest in property. This research focuses on apartment price prediction in real estate market of Belgrade, employing advanced machine learning algorithms, XGBoost and Random Forest, to estimate property values. These algorithms are highly effective at processing large datasets and uncovering complex relationships among key predictors. Data preprocessing plays a crucial role in ensuring model accuracy. This process includes data cleaning, handling missing values, and addressing outliers, while categorical features are transformed through one-hot encoding to enhance model performance. Both XGBoost and Random Forest models are applied to predict apartment prices, providing valuable insights into the most influential factors in the market. A comparative analysis of these methods not only highlights their respective strengths and effectiveness within the context of the Serbian real estate market but also demonstrates their utility in improving prediction accuracy. The results of this study contribute to a more transparent and efficient real estate market in Belgrade, offering valuable tools for all stakeholders: sellers, prospective buyers, policymakers, local administration, etc. By integrating these advanced predictive models, this research also lays the groundwork for future advancements in real estate analytics, presenting a robust framework for market forecasting and strategic decision-making. Although applied to apartment price prediction, the models can be, with minor modifications, applied to price predictions of other types of real estate (houses, commercial, industrial, land plots, etc.). Additionally, the findings suggest potential areas for further research to refine prediction models and adapt them to evolving market conditions.