Comparative analysis of machine learning models in predicting housing prices: a case study of Prishtina's real estate market

Purpose - The purpose of this study is to carry out a comparative analysis of four machine learning models such as linear regression, decision trees, k-nearest neighbors and support vector regression in predicting housing prices in Prishtina. Design/methodology/approach - Using Python, the models were assessed on a data set of 1,512 property transactions with mean squared error, coefficient of determination, mean absolute error and root mean squared error as metrics. The study also conducts variable importance test. Findings - Upon preprocessing and standardization of the data, the models were trained and tested, with the decision tree model producing the best performance. The variable importance test found the distance from central business district and distance to the road leading to central business district as the most relevant drivers of housing prices across all models, with the exception of support vector machine model, which showed minimal importance for all variables. Originality/value - To the best of the author’s knowledge, the originality of this research rests in its methodological approach and emphasis on Prishtina's real estate market, which has never been studied in this context, and its findings may be generalizable to comparable transitional economies with booming real estate sector like Kosovo.