Prediction of profitability determinants in real estate investment trusts through data mining methods: An international comparison

Among real estate types, residential properties a crucial role in meeting people's fundamental needs for shelter, security, and belonging. Real Estate Investment Trusts (REITs) provide an effective tool for gathering, managing, and diversifying real estate investments, offering investors the opportunity to participate in large-scale projects. The study aims to identify the prioritization of financial indicators affecting the profitability of residential real estate investment trust companies and to compare three different data mining methods used for this purpose. In the study, three data mining methods, namely Random Forest Regression, XGBoost, and CatBoost, were employed. The dataset of the study consists of data from the periods 2013.Q1 to 2022.Q1 for 32 residential REIT companies operating predominantly in the residential sector across 7 countries. REIT companies were divided into two groups based on the minimum dividend distribution ratios required for them to benefit from corporate tax exemption and inter-group comparisons were made. The dependent variable used was the interest, depreciation, and tax pre-profit margin. Independent variables consisted of firm size, net asset value, leverage ratio, fixed asset turnover, and current ratio. According to the findings, firm size was identified as the most significant indicator influencing companies' profitability, while the current ratio was found to be the least significant indicator. Additionally, the error values of models created with XGBoost and CatBoost methods were lower than those created with the Random Forest Regression method. This finding indicates that the XGBoost and CatBoost methods are more successful compared to the Random Forest method.