Application of Evolving Fuzzy Systems to Construct Real Estate Prediction Models

摘要

Four variants of the eTS algorithms (evolving Takagi-Sugeno fuzzy systems) were implemented and examined in respect of their usefulness for the intelligent system of real estate market. The eTS algorithms were compared as regards their predictive accuracy with the Flexfis algorithm and ensembles employing general linear models (Glm) devoted to predict from a data stream of real estate sales transactions. The experiments were conducted in Matlab environment using real-world data taken from a dynamically changing real property market. The analysis of the results was performed using statistical methodology including nonparametric tests followed by post-hoc procedures designed especially for multiple N×N comparisons. The models produced by two versions of Simple_eTS and Flexfis algorithms and as well as ensembles composed of Glm models revealed statistically similar performance.