Machine learning models in mass appraisal for property tax purposes: a systematic mapping study

Abstract

The use of machine learning models in mass appraisal of properties for tax purposes has been extensively investigated, generating a growing volume of primary research. This study aims to provide an overview of the machine learning techniques used in this context and analyze their accuracy. We conducted a systematic mapping study to collect studies published in the last seven years that address machine learning methods in the mass appraisal of properties. The search protocols returned 332 studies, of which 22 were selected, highlighting the frequent use of Random Forest and Gradient Boosting models in the last three years. These models, especially Random Forest, have shown predictive superiority over traditional appraisal methods. The measurement of model performance varied among the studies, making it difficult to compare results. However, it was observed that the use of machine learning techniques improves accuracy in mass property appraisals. This article advances the field by summarizing the state of the art in the use of machine learning models for mass appraisal of properties for tax purposes, describing the main models applied, providing a map that classifies, compares, and evaluates the research, and suggesting a research agenda that identifies gaps and directs future studies.