Feature Selection Applications and Model Validation for Mass Real Estate Valuation Systems

Abstract

Real estate valuation, which has great importance in the country’s economy, is one of the issues that cannot be resolved yet. The first of the most important problems in valuation is that the feature that affects the value is not clear, and the second is that the search for methods continues because the classical valuation methods are insufficient. Therefore, mass real estate valuation systems have been revealed. In current laws, the features for valuation are not standard and many different are used in each application, even if the real estate type is the same. Also, since the market conditions in the valuation of real estate are formed according to the supply-demand relationship, the subjective approaches cause the value to affect. In the mass valuation system that needs to be established, it is essential to determine the features that affect the value according to the real estate type and the characteristic of the region to be modeled. In this study, to determine the features needed to be addressed while the mass real estate valuation, a survey application was carried out located in the Central Anatolia region of Turkey Ankara, Konya, and Kayseri province. The survey application was conducted with both experts and citizens. Using survey questions, feature selection was made with Frequency Analysis (FRA), Principal Component Analysis (PCA), Factor Analysis (FA), and Analytical Hierarchy Process (AHP) approach. A total of 21 scenarios were created with participant groups from four different methods and indexing. According to the scenarios obtained from the methods applied as a result of the survey, the Multiple Regression (linear) Analysis method (MRA) was used to examine the verification of the features under market conditions. Three study regions were determined in Konya, and 21 scenario features created were implemented in study regions by the MRA method. The results of the study were examined by applying performance analysis.