Armed conflicts on a global scale have become the focus of the international community,and how to combine multi-source geospatial data research to predict the risk of conflict outbreak from a more refined spatial scale is a problem that needs to be solved.Taking Pakistan as an example,a multi-scale armed conflict prediction model based on XGBoost is proposed.Firstly,a multi-scale grid is con-structed to divide the data,which is used as the thematic dataset of the prediction model after spatio-temporal rasterization.And then 6 thematic sub-models are constructed based on the dataset for geogra-phy,economy and society,time lapse,spatial expansion,spatio-temporal dependence and total ele-ments.Finally,the risk level of the outbreak of armed conflict is classified and predicted by combining the unbalanced sample processing and Bayesian auto-optimization.The XGBoost-based multiscale armed conflict prediction model can achieve finer spatial delineation and accurately capture the loca-tions and distribution patterns of conflict outbreaks.The model can be applied at different levels to pro-vide a reference for the formulation of conflict risk policies in different regions.
万方数据
