Empirical Comparison of Resampling Methods Using Genetic Fuzzy Systems for a Regression Problem

Much attention has been given in machine learning field to the study of numerous resampling techniques during the last fifteen years. In the paper the investigation of m-out-of-n bagging with and without replacement and repeated cross-validation using genetic fuzzy systems is presented. All experiments were conducted with real-world data derived from a cadastral system and registry of real estate transactions. The bagging ensembles created using genetic fuzzy systems revealed prediction accuracy not worse than the experts’ method employed in reality. It confirms that automated valuation models can be successfully utilized to support appraisers’ work.