基于粗集-支持向量机的采空区自然发火预测  

Prediction of spontaneous combustion in caving zone based on rough set and support vector machine

作　　者：孟倩[1] 王永胜[2,3] 周延[2] 

机构地区：[1]徐州师范大学计算科学与技术学院,江苏徐州221116 [2]中国矿业大学安全工程学院,江苏徐州221008 [3]充州煤业股份有限公司通防部,山东邹城273500

出　　处：《煤炭学报》2010年第12期2100-2104,共5页Journal of China Coal Society

基　　金：国家“十一五”科技支撑计划课题(2007baq00168-1);江苏省高校自然科学基础研究资助项目(08KJD520022)

摘　　要：将粗集和支持向量机两种算法有机结合起来,建立了基于粗集与支持向量机的采空区自然发火预测模型。通过粗集对采空区自然发火影响因子进行预处理,将约简属性作为输入向量,利用支持向量机进行分类处理,选择了支持向量机核函数,利用变步长搜索法对支持向量机参数进行了优化。在对粗集-支持向量机方法的实验中,通过与支持向量机方法和神经网络方法的比较,发现在样本有限的情况下,基于粗集-支持向量机的采空区自然发火预测方法预测精度更高,训练速度更快,为采空区自然发火预测提供了一种新的方法。A approach to predict spontaneous combustion in caving zone by using rough set and support vector machine（RS-SVM） was proposed.The original sample data were preprocessed with the knowledge reduction algorithm of RS theory,and the redundant condition attributes and conflicting samples were eliminated from the working sample set to reduce space dimension of sample data.Preprocessed sample data were used as training sample data of SVM.With choosing an appropriate kernel function,the SVM parameters were optimized by using variable step search algorithm.Finally,a comparison of the performance of RS-SVM with SVM and neural networks was carried out.The experimental results show that the prediction based on RS-SVM can improve the training speed and precision of classification when the samples are limited.

关 键 词：粗集-支持向量机 采空区自然发火 预测 神经网络 

分 类 号：TD752.2[矿业工程—矿井通风与安全]