矿区采空塌陷危险性预测的RS-SVM模型  

RS-SVM model for predicting underground goaf collapse risk

作　　者：温廷新[1] 孙红娟[1] 徐波[1] 邵良杉[1] 章菲菲[1] 

机构地区：[1]辽宁工程技术大学系统工程研究所

出　　处：《中国安全科学学报》2015年第10期16-21,共6页China Safety Science Journal

基　　金：国家自然科学基金资助(71371091);辽宁省教育厅基金资助(L14BTJ004);山东省自然科学基金资助(ZR2010FL012)

摘　　要：为快速、准确地预测矿区采空塌陷的危险性,针对矿区采空塌陷预测的复杂非线性特点,在统计分析实测资料的基础上,选取7项指标作为初始特征指标,30组塌陷样本作为原始样本,其中,前17组为原始训练样本,后13组为测试样本;运用粗糙集(RS)理论,对原始训练样本进行对象约简和属性约简。将属性约简后的3项指标作为支持向量机(SVM)的输入向量,建立矿区采空塌陷危险性预测的RS-SVM模型。将对象约简后的7组样本作为训练样本,用于模型训练。采用回代估计法进行回检,误判率为0。利用训练好的模型对13组待评样本进行预测,并与贝叶斯、BP神经网络(BPNN)方法进行比较。结果表明,RS理论与SVM算法相结合,能降低属性维数,去除冗余样本,简化模型,该模型所得预测结果准确率为100%。In order to predict the risk of mining collapse accurately and fast,for the complex nonlinear characteristics of predicting the underground goaf collapse risk,on the basis of statistical analysis of historical collapse information,seven indexes were selected as the initial characteristic indexes,and 30 groups of collapse samples were taken as the original samples,the first 17 groups as the original training samples,and the other 13 groups as the test samples. By using RS theory,the original training samples were reduced in both object and attribute. 3 indexes after attribute reduction were taken as the input vectors of SVM. An RS-SVM model was built for predicting the underground goaf collapse risk. The 7 grounds of training samples after object reduction were used for training the model. The re-substitution method was used for verifying the model and the misjudgment rate was 0. The trained model was used for predicting the13 groups of test samples. A prediction result comparison was made between the model and the methods of Bayes and BPNN. The results show that the combination of RS theory and SVM algorithm can reduce the attribute dimensions,remove the redundant samples and simplify the model,and that the prediction accuracy of the model is 100%.

关 键 词：采空区 塌陷危险性预测 粗糙集(RS)理论 支持向量机(SVM) 属性约简 

分 类 号：X936[环境科学与工程—安全科学] TD327[矿业工程—矿井建设]