基于随机森林的山洪灾害风险评价方法及应用

Mountain Torrent Disaster Risk Assessment Method and Application Based on Random Forest

针对空间尺度较小区域的山洪灾害风险评价中指标权重难以准确确定的问题,选取林州市为研究对象,采用后果逆向扩散法从自然灾害系统和社会灾害系统2个方面选取了9个指标构建山洪灾害风险指标体系;引入随机森林算法,建立林州市随机森林风险评价模型,根据不同山洪灾害风险等级划分林州市山洪灾害风险区,并与历史山洪灾害发生点的实际情况进行对比分析。结果表明:随机森林风险评价模型的风险区等级划分精度达到81.34%,证明随机森林算法运用于空间尺度较小区域的山洪灾害风险评价时具有较高的准确性。

Aiming at the issue that it was difficult to accurately determine the index weight in the risk assessment of mountain torrent disasters in regions with small spatial scale,this paper selected Linzhou City as the research object,adopted the consequence reverse diffusion meth⁃od,and selected 9 indicators from 2 aspects of natural disaster system and social disaster system to build a mountain torrent disaster risk in⁃dex system.The random forest algorithm was introduced to establish the random forest risk assessment model of Linzhou City,according to different mountain torrent disaster risk levels,the mountain torrent disaster risk areas in Linzhou City were divided,and the comparison and analysis were made with the actual situation of historical mountain torrent disaster occurrence points.The results show that the classification accuracy of risk areas of the random forest risk assessment model reaches to 81.34%,which proves that the random forest algorithm has high accuracy when is applied to the risk assessment of mountain torrent disasters in areas with small spatial scales.