Determining sub-real estate markets with hybrid gradual unsupervised learning for better real estate price prediction performance

This study, various clustering methods that enable the determination of data-based housing submarkets, which is a subfield of unsupervised learning, were used to determine the most suitable market segments in order to improve the performance of real estate price prediction modeling. In order to preserve spatial contiguity, this study first performs a priori classification and ad hoc subdivision according to real estate size, and then performs clustering analysis only on latitude and longitude spatial features. In addition, the empirical findings obtained within the scope of the study show that such an approach produces more consistent results than cluster analyzes performed with alternative feature sets.