Comparative Assessment of Radom Forest, SVC and Cat Boost Performances as Property Price Forecasting Models

Abstract

Associated risk and uncertainties that characterized property investment necessitate adoption of property price forecasting models with high precision rate. This study aimed at comparing the predictive performances of the random forest, support vector machine, cat boost as residential property price forecasting models. Data for the study were gathered from the records of recent lettings as provided by residential property managers in the study area. For the purpose of precision, this study adopted random forest, SVM and Cat Boost as model of classifying rental value of residential properties in the study area. Important property value determining factors like; distant to cultural site, age of building, house type, road network, availability of water, state of exterior, state of interior and security were considered as input variables. It was revealed that the three adopted forecasting models achieved over 80% of precision and accuracy in the classification of residential properties in the study area. Also, the study established that random forest as the best rental value forecasting model among the three considered rental value forecasting models.