改进GSM-RFC模型在回采巷道围岩稳定性分级的预测
|
摘要
为对回采巷道围岩稳定性等级进行准确分类,在分析回采巷道围岩稳定性影响因素的基础上,采用改进网格搜索法(GSM)对随机森林分类(RFC)算法关键参数进行搜索确定.首先对RFC中生成决策树的叶节点最小记录百分比进行最优选值,而后以预测准确率为目标函数,借助改进GSM两次搜索确定RFC关键参数,并对各影响因素重要程度进行排序.从95组现场数据中选取80组作为训练集,15组为测试集,并将预测结果与GSM-RFC、RFC对比.研究结果表明:RFC最优叶节点最小记录百分比为58%,最优分裂属性值为3,最优决策树棵树为420;较GSM-RFC与RFC模型,改进GSM-RFC模型有更高的准确率(97.778%)、Kappa系数(0.970)和较合理的运行时间(482.772s),表明改进GSM-RFC模型具有更好的拟合效果和泛化误差,可以满足工程实际需要.
In order to classify the surrounding rocks stability classification of gateway accurately,based on the analysis of the factors influencing the stability of surrounding rocks,this paper utilized an improved grid search method(GSM) to search and select the key parameter of random forest classification(RFC) algorithm.Firstly,this study optimized the minimum percentage of the leaf nodes which control decision tree generation in RFC,then took the prediction accuracy as the objective function,determined the key parameters of RFC by means of used the improved GSM two search,and sorted the influence factors of the stability of the surrounding rock of gateway.Through practical investigation,95 sets of data were obtained,80 groups selected as training data sets and 15 group as test data set,and the prediction results are compared with the GSM-RFC and RFC algorithms.The results show that the optimal minimum percentage of leaf nodes is 58%,the optimal split attribute value is 3,and the optimal decision tree is 420;compared with GSM-RFC and RFC model,the improved GSM-RFC model has higher accuracy(97.778%),Kappa coefficient(0.970) and reasonable running time(482.772 s).This study shows that the improved GSM-RFC model has better fitting effect and generalization error,and can meet the actual needs of the project.
In order to classify the surrounding rocks stability classification of gateway accurately,based on the analysis of the factors influencing the stability of surrounding rocks,this paper utilized an improved grid search method(GSM) to search and select the key parameter of random forest classification(RFC) algorithm.Firstly,this study optimized the minimum percentage of the leaf nodes which control decision tree generation in RFC,then took the prediction accuracy as the objective function,determined the key parameters of RFC by means of used the improved GSM two search,and sorted the influence factors of the stability of the surrounding rock of gateway.Through practical investigation,95 sets of data were obtained,80 groups selected as training data sets and 15 group as test data set,and the prediction results are compared with the GSM-RFC and RFC algorithms.The results show that the optimal minimum percentage of leaf nodes is 58%,the optimal split attribute value is 3,and the optimal decision tree is 420;compared with GSM-RFC and RFC model,the improved GSM-RFC model has higher accuracy(97.778%),Kappa coefficient(0.970) and reasonable running time(482.772 s).This study shows that the improved GSM-RFC model has better fitting effect and generalization error,and can meet the actual needs of the project.
引文
[1]王琦.乌海矿区回采巷道围岩稳定性分类及其控制[D].徐州:中国矿业大学,2015.
[2]周保生,朱维申,李术才.综放回采巷道围岩稳定性分类的研究[J].煤炭学报,2000(5):469-472.ZHOU Baosheng,ZHU Weishen,LI Shucai.Stability classification of mining roadway for fully mechanized sublevel caving face[J].Journal of China Coal Society,2000(5):469-472.
[3]朱一丁,马文涛.回采巷道围岩分类的支持向量机方法[J].采矿与安全工程学报,2006(3):362-365.ZHU Yiding,MA Wentao.Method of support vector machines for classifying surrounding rocks of gate-ways[J].Journal of Mining&Safety Engineering,2006(3):362-365.
[4]孔祥松,单仁亮,肖禹航.基于SPSS与ANN的汾西矿区煤巷围岩稳定性分类[J].矿业安全与环保,2016,43(3):13-17.KONG Xiangsong,SHAN Renliang,XIAO Yuhang.Cl-assification of surrounding rock stability of coal road-ways in Fenxi coal mining area based on SPSS and ANN[J].Mining Safety&Environmental Protection,2016,43(3):13-17.
[5]李春萍,郝会兵.煤巷围岩分类的Bayes判别分析法[J].煤炭学报,2011,36(S2):304-307.LI Chunping,HAO Huibing.Method of Bayes discrimin-ant analysis for classifying mine-lane surrounding rocks[J].Journal of China Coal Society,2011,36(S2):304-307.
[6]刘洋,叶义成,刘晓云.基于未确知聚类法的巷道围岩稳定性预测[J].中国安全生产科学技术,2017,13(2):56-61.LIU Yang,YE Yicheng,LIU Xiaoyun.Stability forecast on surrounding rock of roadway based on unascertained clustering method[J].Journal of Safety Science and Technology,2017,13(2):56-61.
[7]汪明武,陈光怡,金菊良.基于多元联系数-三角模糊数随机模拟的围岩稳定性风险评价[J].岩土工程学报,2011,33(4):643-647.WANG Mingwu,CHEN Guangyi,JIN Juliang.Risk eva-luation of surrounding rock stability based on stochastic simulation of multi-element connection number and traingular fuzzy numbers[J].Chinese Journal of Geotechnical Engineering,2011,33(4):643-647.
[8]LIAW A,WIENER M.Classification and regression with random forest[J].R News,2002,2(3):18-22.
[9]曹正凤.随机森林算法优化研究[D].北京:首都经济贸易大学,2014.
[10]邓生雄,雒江涛,刘勇.集成随机森林的分类模型[J].计算机应用研究,2015,32(6):1 621-1 624,1 629.DENG Shengxiong,LUO Jiangtao,LIU Yong.Classific-ation model based on ensemble random forest[J].Application Research of Computers,2015,32(6):1 621-1 624,1 629.
[11]郭超,宋卫华,魏威.基于网格搜索-支持向量机的采场顶板稳定性预测[J].中国安全科学学报,2014,24(8):31-36.GUO Chao,SONG Weihua,WEI Wei.Stope roof stability prediction based on both SVM and grid-search method[J].China Safety Science Journal,2014,24(8):31-36.
[12]王健峰.基于改进网格搜索法SVM参数优化的说话人识别研究[D].哈尔滨:哈尔滨工程大学,2012.
[13]奉国和,朱思铭.基于聚类的大样本支持向量机研究[J].计算机科学,2006(4):145-147.FENG Guohe,ZHU Siming.Clustering-based large scale sample support vector machines[J].Computer Science,2006(4):145-147.
[14]袁颖,王晨晖,周爱红.巷道围岩稳定性分类的GSM-SVM预测模型[J].煤矿安全,2017,48(6):200-203.YUAN Ying,WANG Chenhui,ZHOU Aihong.Prediction model for stability classification of roadway surrounding rock based on grid search method and support vec-tor machine[J].Safety in Coal Mines,2017,48(6):200-203.
[15]贺超峰,华心祝,马菁花.基于BP神经网络的回采巷道围岩稳定性分类[J].矿业工程研究,2012,27(3):6-9.HE Chaofeng,HUA Xinzhu,MA Jinghua.Classification of surrounding rock stability of roadway based on BP neural network[J].Mineral Engineering Research,2012,27(3):6-9.
[16]司康平,田原,汪大明.滑坡灾害危险性评价的3种统计方法比较-以深圳市为例[J].北京大学学报(自然科学版),2009,45(4):19-26.SI Kangping,TIAN Yuan,WANG Daming.The compareson of three statistical methods on landslide susceptibility analysis:a case study of Shenzhen City[J].Acta Scientiarum Naturalium Universitatis Pekinensis,2009,45(4):19-26.
[2]周保生,朱维申,李术才.综放回采巷道围岩稳定性分类的研究[J].煤炭学报,2000(5):469-472.ZHOU Baosheng,ZHU Weishen,LI Shucai.Stability classification of mining roadway for fully mechanized sublevel caving face[J].Journal of China Coal Society,2000(5):469-472.
[3]朱一丁,马文涛.回采巷道围岩分类的支持向量机方法[J].采矿与安全工程学报,2006(3):362-365.ZHU Yiding,MA Wentao.Method of support vector machines for classifying surrounding rocks of gate-ways[J].Journal of Mining&Safety Engineering,2006(3):362-365.
[4]孔祥松,单仁亮,肖禹航.基于SPSS与ANN的汾西矿区煤巷围岩稳定性分类[J].矿业安全与环保,2016,43(3):13-17.KONG Xiangsong,SHAN Renliang,XIAO Yuhang.Cl-assification of surrounding rock stability of coal road-ways in Fenxi coal mining area based on SPSS and ANN[J].Mining Safety&Environmental Protection,2016,43(3):13-17.
[5]李春萍,郝会兵.煤巷围岩分类的Bayes判别分析法[J].煤炭学报,2011,36(S2):304-307.LI Chunping,HAO Huibing.Method of Bayes discrimin-ant analysis for classifying mine-lane surrounding rocks[J].Journal of China Coal Society,2011,36(S2):304-307.
[6]刘洋,叶义成,刘晓云.基于未确知聚类法的巷道围岩稳定性预测[J].中国安全生产科学技术,2017,13(2):56-61.LIU Yang,YE Yicheng,LIU Xiaoyun.Stability forecast on surrounding rock of roadway based on unascertained clustering method[J].Journal of Safety Science and Technology,2017,13(2):56-61.
[7]汪明武,陈光怡,金菊良.基于多元联系数-三角模糊数随机模拟的围岩稳定性风险评价[J].岩土工程学报,2011,33(4):643-647.WANG Mingwu,CHEN Guangyi,JIN Juliang.Risk eva-luation of surrounding rock stability based on stochastic simulation of multi-element connection number and traingular fuzzy numbers[J].Chinese Journal of Geotechnical Engineering,2011,33(4):643-647.
[8]LIAW A,WIENER M.Classification and regression with random forest[J].R News,2002,2(3):18-22.
[9]曹正凤.随机森林算法优化研究[D].北京:首都经济贸易大学,2014.
[10]邓生雄,雒江涛,刘勇.集成随机森林的分类模型[J].计算机应用研究,2015,32(6):1 621-1 624,1 629.DENG Shengxiong,LUO Jiangtao,LIU Yong.Classific-ation model based on ensemble random forest[J].Application Research of Computers,2015,32(6):1 621-1 624,1 629.
[11]郭超,宋卫华,魏威.基于网格搜索-支持向量机的采场顶板稳定性预测[J].中国安全科学学报,2014,24(8):31-36.GUO Chao,SONG Weihua,WEI Wei.Stope roof stability prediction based on both SVM and grid-search method[J].China Safety Science Journal,2014,24(8):31-36.
[12]王健峰.基于改进网格搜索法SVM参数优化的说话人识别研究[D].哈尔滨:哈尔滨工程大学,2012.
[13]奉国和,朱思铭.基于聚类的大样本支持向量机研究[J].计算机科学,2006(4):145-147.FENG Guohe,ZHU Siming.Clustering-based large scale sample support vector machines[J].Computer Science,2006(4):145-147.
[14]袁颖,王晨晖,周爱红.巷道围岩稳定性分类的GSM-SVM预测模型[J].煤矿安全,2017,48(6):200-203.YUAN Ying,WANG Chenhui,ZHOU Aihong.Prediction model for stability classification of roadway surrounding rock based on grid search method and support vec-tor machine[J].Safety in Coal Mines,2017,48(6):200-203.
[15]贺超峰,华心祝,马菁花.基于BP神经网络的回采巷道围岩稳定性分类[J].矿业工程研究,2012,27(3):6-9.HE Chaofeng,HUA Xinzhu,MA Jinghua.Classification of surrounding rock stability of roadway based on BP neural network[J].Mineral Engineering Research,2012,27(3):6-9.
[16]司康平,田原,汪大明.滑坡灾害危险性评价的3种统计方法比较-以深圳市为例[J].北京大学学报(自然科学版),2009,45(4):19-26.SI Kangping,TIAN Yuan,WANG Daming.The compareson of three statistical methods on landslide susceptibility analysis:a case study of Shenzhen City[J].Acta Scientiarum Naturalium Universitatis Pekinensis,2009,45(4):19-26.