Land price prediction model for housing development in metropolitan area using deep learning technique: A case study of Thailand

Abstract

Thailand has experienced a significant concentration of development in Bangkok, leading to a continual and substantial increase in land prices within the capital city. This phenomenon has driven greater interest among investors in the metropolitan areas surrounding Bangkok, which are less dense but show promising potential for future development. However, investing in these new areas requires extensive knowledge, the experience of professional appraisers, and highly accurate data. Therefore, this research aimed to propose a method for analyzing land value in the Bangkok metropolitan region using Deep Learning technique. The goal was to offer real estate developers a more precise and reliable tool for evaluating appropriate land prices. The research methodology included the collection of vacant land data within Bangkok and its surrounding provinces from feasyonline.com, a credible real estate data source in Thailand. The data was then analyzed using Deep Learning technique, considering 29 independent variables that influence land prices. These variables can be grouped into five key factors, ranked by importance: (1) type of business in the area, (2) infrastructure, utilities, and community services, (3) specific physical characteristics of the land, (4) legal and regulatory constraints, and (5) location-related factors. The model developed in this study demonstrated high performance, with a Coefficient of Determination (R2) of 0.71 and a Root Mean Square Error (RMSE) of 6,068.08—both considered acceptable values and better than those of the application’s Auto Model. The research findings can be applied in two main ways. First, in a business context, investors and developers can use the model to support decision-making when acquiring lands for new projects, designing project types, and determining appropriate selling prices. Second, in academic development, researchers and interested individuals can adapt the Deep Learning technique for studies involving real estate business analysis with limited data, or customize the database, project types, or incorporate additional variables into the model.