Predicting apartment prices in Prishtina: insights from gradient boosting modeling

Purpose This paper aims to apply gradient boosting (GB) to forecast apartment prices in Prishtina, evaluating its effectiveness through established evaluation metrics and identifying key price determinants via variable importance analysis. Design/methodology/approach The research method comprises analyzing a dataset of 1,468 property transactions from 2019 to 2023, attained from the Kosovo Department of Property Taxes. The data were split into training and testing subsets, with normalization applied to provide consistency, and the GB model's performance was evaluated using Mean Squared Error, Coefficient of Determination, Mean Absolute Error and Root Mean Squared Error. Findings Findings point out that the GB model shows noteworthy predictive accuracy, with the distance from the Central Business District identified as the primary influencer of apartment prices in Prishtina. These findings also point out the model's capability to effectively capture and reflect the nuances of the local real estate market. Practical implications Implications of this study advise the practical utility of machine learning, specifically GB, in enhancing real estate valuation and decision-making processes. By providing accurate and efficient market analyses, such models can enable more informed investment decisions, contribute to market transparency and help in stabilizing property prices. Originality/value Originality of this research lays in its focused application of Gradient Boosting within an emerging market context, enriching the empirical discourse on real estate valuation. By conducting a comprehensive variable importance analysis, the study offers fresh understanding into the dynamics influencing property prices, thereby advancing the field of real estate research.