Analyzing Mass Appraisal of Urban Residential Land with Machine Learning - A Case Study in Hanoi, Vietnam

As the real estate market expands, the necessity for swift and accurate land appraisal within specific market conditions becomes increasingly pertinent and socially significant. Among various methods for land appraisal, machine learning emerges as a novel approach, and the number of machine learning algorithms has rapidly increased over the past decade. Consequently, it is necessary to conduct comparative studies on the performance of various machine learning algorithms for a mass appraisal of land. The objective of this study is to evaluate and compare five machine learning algorithms, such as Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost). The dataset for training and testing comprises 1082 observations spanning 12 districts of Hanoi City, encompassing 20 features related to physical, locational, legal, and social properties of the land. The results show that RF and GB are, overall, the best algorithms in terms of accuracy metrics and correspondence of feature importance for land price with correlation analysis. The locational features of land parcels (Location, Zone, CityCenter) and street features (RoadWidth, RoadType) have the biggest impact on land price. These findings underscore the potential of machine learning applications in the mass appraisal of urban residential land, primarily when supported by a substantial training dataset.