Recommendation of Regression Models for Real Estate Price Prediction using Multi-Criteria Decision Making

Abstract

Accurate prediction of real estate prices is an essential task for establishing real estate policies. Even though various regression models for real estate price prediction have been developed so far, selecting the most suitable regression model is a challenging task since the performance of different regression models varies for different accuracy measures. This paper aims to recommend the most suitable regression model for real estate price prediction, considering various performance measures altogether using multi-criteria decision making (MCDM). The evaluation of regression models involves a number of competing accuracy measures; hence, choosing the best regression model for predicting real estate price is modeled as the MCDM problem in the proposed approach. An experimental study is designed using 22 regression models, three MCDM methods, six performance measures, and three real estate price datasets to validate the proposed approach. Experimental outcomes show that Gradient Boosting, Random Forest, and Ridge Regression are recommended as the best regression models based on MCDM ranking. The results of the experimental study show that the proposed MCDM-based strategy can be utilized effectively in real estate industries to choose the best regression model for predicting real estate prices by optimizing several competing accuracy measures.