A Study on Identification of Urban Waterlogging Risk Factors Based on Satellite Image Semantic Segmentation and XGBoost

As global warming exacerbates and urbanization accelerates, extreme climatic events occur frequently. Urban waterlogging is seriously spreading in China, resulting a high level of vulnerability urban societies economies. It has been urgent for regional sustainable development to effectively identify analyze the risk factors behind waterlogging. A novel model incorporating satellite image semantic segmentation into gradient boosting (XGBoost) employed identifying forecasting factors. Ground object features points are extracted by segmentation, XGBoost predict primary affecting This paper selects coastal cities Haikou, Xiamen, Shanghai, Qingdao as research areas, obtains data from social media. According comprehensive performance evaluation models, could extract water bodies, roads, green spaces images, more accurate reliable than other common machine learning methods prediction precision. Among all factors, elevation main factor areas. For Shanghai Qingdao, secondary roads. Water bodies Haikou. four indicators equally significant, which be regarded