Prediction of Land Price for Sustainable Housing Development in the Capital of Thailand Using Deep Learning Techniques

Due to the high population density and limited land availability in Bangkok, the capital of Thailand, land values have been increasing every year, posing challenges to sustainable housing development. Accurate land valuation is critical not only for investment decisions but also for promoting economic efficiency, social equity, and sustainable urban land use. Inaccurate analysis can lead to losses for real estate developers, project residents, and surrounding communities. However, this process requires extensive knowledge and experience. This research presents an approach for analyzing land values in Bangkok using Deep Learning techniques, which can help real estate developers assess appropriate land values more accurately and precisely. The study collected data on vacant land in Bangkok from an online feasibility study database and analyzed them using Deep Learning techniques. The results showed 30 determinants categorized into five groups. The study conducted 80 parameter adjustments with a ratio of 128:64:32 using a Quadratic Loss Function. The model was validated using k-fold cross-validation to ensure robustness and a Model Simulator operator to test sensitivity analysis. The Deep Learning model resulted in an R-square value of 0.917 and an RMSE of 2620 USD. The results of this research can be used as an effective decision-making tool for real estate developers, landowners, and brokers in determining appropriate buying or selling prices for land to support real estate sustainable development.