摘要
针对LSSVM参数优化耗时长且效果差的问题,提出了一种基于差分蝙蝠算法(DE-BA)的LSSVM参数寻优办法,利用BA的全局寻优能力和DE的局部搜索能力,优化LSSVM的核参数(C)和惩罚参数(σ2),并建立了基于DEBA-LSSVM的露天矿边坡稳定性预测模型,选取某露天矿边坡实际数据进行了对比分析。结果表明:3种预测模型中,DE-BA-LSSVM模型的平均相对误差为2.6%,预测效果最好,为采用LSSVM解决露天矿边坡稳定性预测问题提供了新的方向。
The parameter optimization of LSSVM takes a long time and has a poor result. Aimed at these problems, a LSSVM parameter optimization method based on differential bat algorithm （DE-BA） was proposed. By using the global optimization a-bility of BA and local search ability of DE, the kernel parameters （C） and penalty parameters （ az ） of LSSVM were optimized, and the slope stability prediction model of open-pit mine was established based on DE-BA-LSSVM. Then, the actual data of an open-pit slope was selected to compare and analyze. The results showed that the average relative error of DE-BA-LSSVM model was 2.6%, with the best prediction effect in the three forecasting models. Therefore, the DE-BA-LSSVM model provided a new direction to use LSSVM to solve the prediction problems of slope stability in open-pit mine.
作者
顾清华
李梦然
闫宝霞
GU Qinghua;LI Mengran;YAN Baoxia(Xi＇an University of Architecture and Technology,Xi＇an,Shaanxi 710055,China;Xi＇an Engineering and Research Institute of Nonferrous Metallurgy Limited Company,Xi＇an,Shaanxi 710001,China)
出处
《矿业研究与开发》
CAS
北大核心
2018年第8期1-5,共5页
Mining Research and Development
基金
国家自然科学基金项目(51774228)
陕西省自然科学基金项目(2017JM5043)
关键词
差分蝙蝠算法
最小二乘支持向量机
边坡稳定性
预测模型
Differential bat algorithm
Least squares support vector machine
Slope stability
Prediction model
