基于RF-ELM模型的边坡稳定性预测研究  

Study on slope stability prediction based on RF- ELM model

作　　者：邵良杉[1] 马寒[1] 温廷新[1] 

机构地区：[1]辽宁工程技术大学系统工程研究所,辽宁葫芦岛125105

出　　处：《中国安全生产科学技术》2015年第3期93-98,共6页Journal of Safety Science and Technology

基　　金：国家自然科学基金资助项目(71371091);辽宁省高等学校杰出青年学者成长计划项目(LJQ2012027)

摘　　要：为实现对边坡稳定性的有效预测,将极限学习机算法与旋转森林算法相结合,并依据影响边坡稳定性的六项重要因素,建立了边坡稳定性预测的RF-ELM预测模型。该模型是以极限学习机算法为基分类器,以旋转森林算法为框架的集成学习模型,利用UCI数据库中三组数据集验证了该集成模型确实提高了ELM的预测性能。将RF-ELM模型应用于边坡稳定性的预测问题中,结合39组工程实例数据进行预测实验,结果表明该模型具有较高的预测精度,可有效的对边坡稳定性进行预测。In order to predict the slope stability effectively , considering the six important influence factors of slope stability , a RF-ELM forecasting model of slope stability was established by combining extreme learning algorithm and rotation forest algorithm .This model is an integrated learning method , which uses extreme learning algorithm as base classifier and rotation forest algorithm as integration framework .A prediction test on 3 data sets of UCI database proved that the model can improve the prediction performance of ELM .By applying RF-ELM model in slope engineering , the prediction experiments were conducted on 39 groups of data in engineering cases .The results showed that RF-ELM model has a higher forecasting accuracy , and it is an effective model for predicting slope staility correctly.

关 键 词：边坡稳定性 极限学习机 旋转森林 分类器集成 

分 类 号：X936[环境科学与工程—安全科学]