基于残缺数据的煤与瓦斯突出预测系统  

Coal and gas outburst prediction system based on incomplete data

作　　者：汝彦冬 吕兴凤[2] 郭继坤 张洪全 陈丽娟 RU Yan-dong;LYV Xing-feng;GUO Ji-kun;ZHANG Hong-quan;CHEN Li-juan(School of Electronics and Information Engineering,Heilongjiang University of Science and Technology,Harbin 150027,China;School of Computer Science,Heilongjiang University,Harbin 150080,China)

机构地区：[1]黑龙江科技大学电子与信息工程学院,哈尔滨150027 [2]黑龙江大学计算机学院,哈尔滨150080

出　　处：《哈尔滨商业大学学报（自然科学版）》2021年第2期172-176,共5页Journal of Harbin University of Commerce：Natural Sciences Edition

基　　金：黑龙江省省属高等学校基本科研业务费项目(No.Hkdqg201909)。

摘　　要：煤与瓦斯突出是煤矿事故发生的主要原因.正确预测煤与瓦斯突出,对于煤炭企业安全生产具有重要意义.对和煤与瓦斯突出相关的5个特征进行分析,采用拉依达准则处理数据异常值.针对数据变量缺失的情况,选择具有代表性的均值插补、多重插补、K近邻插补和随机森林插补方法完成数据插补.采用随机森林、支持向量机和K近邻模型进行煤与瓦斯突出预测.采用精确度,特异度和敏感度作为性能指标.实验表明,采用随机森林数据插补方法并利用随机森林模型完成的预测,取得了精度98.94%、敏感度98.67和特异度100%的性能,在所有数据插补方法和预测模型组合中性能最优.Coal and gas outburst is the main cause of coal mine accidents.Correct prediction of coal and gas outburst is of great significance for the safety production of coal enterprises.In this paper,five characteristic data related to coal and gas outburst were analyzed,and the data outliers were processed by using pauta criterion.In view of the missing data variables,the representative mean value filling,multiple value filling,K nearest neighbor filling and random forest filling were selected to complete the data filling.Random forest,support vector machine and K nearest neighbor model were used to predict coal and gas outburst.Accuracy,specificity and sensitivity were used as performance indicators.The results showed that the accuracy,sensitivity and specificity are 98.94%,98.67%and 100%respectively using the random forest data filling method and random forest prediction model,which were the best in all data filling methods and prediction models.

关 键 词：煤与瓦斯突出 煤矿安全 数据插补 机器学习 拉依达准则 随机森林 数据质量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]