Computational models for comparing spatial data forecast

Spatial data indicate the geographic location of various phenomena, which can be applied in many areas, such as urban expansion analysis or resource use planning. This study uses two groups of spatial data: weather data that affect solar irradiance and data on factors that affect the price of vacant land in Phra Nakhon Si Ayutthaya Province, using the eXtreme Gradient Boosting (XGBoost) and Genetic Algorithm with eXtreme Gradient Boosting (GA- XGBoost) models for forecasting. In particular, incorporating the Genetic Algorithm (GA) with the machine learning model improves the efficiency of finding the appropriate parameter values and better supports the complexity of the spatial data. The experimental results show that XGBoost with seasonal variables provides the best solar intensity forecasting results (???????????????? = 80.36, ????2 = 0.9231) while eliminating seasonal variables significantly reduces the accuracy. Inland price forecasting, using XGBoost combined with data imputation by Multiple Linear Regression (MLR) can explain up to 98% of the price variance. In addition, the attractiveness of the land investment is analyzed and displayed on a GIS map to support informed investment decisions.