Prediction Project - Machine Learning

Hello, I need a machine learning project, more precisely using Randon Forest Spatial Global + Randon Forest Spatial Local (Geographic) + Gradient Boosting + XGBosting + Light Bostinh to predict the value of real estate as a function of a dataset with dozens of explanatory variables , including latitude and lomgitude coordinates. Ideally, the models determine which variables are important and select the 15 best ones for the next step, for example. In the end, it is desired that a comparative table containing the main evaluation metrics (RMSE, MAE, MAPE, R3, MSE, ERROR, COD, PRD) be presented. It is hoped that explanatory graphs and maps are presented for a better understanding of the phenomenon. I have the dataset with the data for simulation and the shapfiles for spatial analysis. It is desirable that the code be written in R language.