基于粗糙集和BP神经网络的滑坡易发性评价

发布时间：2018-05-17 11:28

本文选题：滑坡易发性评价 + 因子筛选　；参考：《煤田地质与勘探》2017年06期

【摘要】：区域滑坡易发性评价是国土规划和滑坡中长期防治的重要依据。为进一步提高滑坡易发性评价的准确性,以恩施市龙凤镇为研究区,运用地理信息系统GIS技术,获取了包括工程岩组、坡度、地质构造等在内的13个初始评价因子,利用基于遗传约简算法的粗糙集理论对初始评价因子进行属性约简,去掉冗余属性后获得最小约简,即8个核评价因子:工程岩组、高程、地形曲率、道路、水系、坡度、坡向、径流强度指数,并以此作为BP神经网络的输入层,构建RS-BPNN预测模型,获得滑坡易发性指数LSI及滑坡易发性等级分区图。其中高易发区面积占总面积的12.82%,该区包含的滑坡面积占总滑坡面积的78.11%,通过ROC曲线测试,模型预测精度为90.9%。结果表明,RS-BPNN模型预测性能良好,进一步提高了滑坡易发性评价的精度和准确性,有较高的工程实用价值。
[Abstract]:Regional landslide vulnerability evaluation is an important basis for land planning and long-term landslide prevention. In order to further improve the accuracy of landslide vulnerability evaluation, 13 initial evaluation factors, including engineering rock group, slope, geological structure and so on, were obtained by using GIS technology in Longfeng Town, Enshi City. The rough set theory based on genetic reduction algorithm is used to reduce the attribute of the initial evaluation factor, and the minimum reduction is obtained by removing the redundant attributes, that is, eight core evaluation factors: engineering rock group, elevation, topographic curvature, road, water system, slope, etc. Slope direction, runoff intensity index, and as the input layer of BP neural network, the RS-BPNN prediction model is constructed, and the landslide susceptibility index LSI and landslide susceptibility grade zoning map are obtained. The area of high risk area accounts for 12.82% of the total area, and the area of landslide contains 78.11% of the total area of landslide. Through the ROC curve test, the prediction accuracy of the model is 90.9. The results show that the prediction performance of RS-BPNN model is good, and the accuracy and accuracy of landslide susceptibility evaluation are further improved.
【作者单位】：中国地质大学(武汉);贵阳学院;
【分类号】：P642.22

【相似文献】