Experimental Comparison of Resampling Methods in a Multi-Agent System to Assist with Property Valuation

A new version of the a multi-agent system to aid in real estate appraisal, called MAREA-2, was introduced. The system is being developed using Java Spring Framework and is intended for industrial application in cadastral information centres. The major part of the study was devoted to investigate the performance of Bagging, Subagging, and Repeated cross-validation models. The overall result of our investigation was that the majority of models created using resampling techniques provided better or equivalent accuracy than the experts' method employed in reality. It confirms that automated valuation models can be successfully incorporated into the multi-agent system and be utilized to support appraisers' work.