融合特征选取与机器学习的煤矿安全生产态势预测  

Coal Mine Safety Production Situation Prediction Based on Feature Selection and Machine Learning

作　　者：叶黎明 施式亮 鲁义[1] 李贺 YE Liming;SHI Shiliang;LU Yi;LI He(School of Resources,Environment and Safety Engineering,Hunan University of Science and Technology,Xiangtan 411201,China;Geologic Team No.294 Nuclear Industry of Fujian Province,Fuzhou 350019,China;Hunan Provincial Key Laboratory of Safe Mining Techniques of Coal Mines,Hunan University of Science and Technology,Xiangtan 411201,China)

机构地区：[1]湖南科技大学资源环境与安全工程学院,湖南湘潭411201 [2]福建省核工业二九四大队,福建福州350019 [3]湖南科技大学煤矿安全开采技术湖南省重点实验室,湖南湘潭411201

出　　处：《湖南科技大学学报(自然科学版)》2024年第4期28-36,共9页Journal of Hunan University of Science And Technology:Natural Science Edition

基　　金：国家自然科学基金资助项目(51974120,51774135)。

摘　　要：为提高煤矿安全态势的预测精度,提出一种基于特征选取与机器学习融合的煤矿安全生产态势预测模型.首先,对1978年—2019年我国煤矿安全生产态势的相关数据进行归一化处理,使用全子集回归和随机森林2种特征选择方法对8种煤矿安全生产态势的影响因素进行组合;然后,运用岭回归、分类与回归树、自适应提升和极端梯度提升4种机器学习算法分别对12种煤矿安全生产态势特征参数组合进行预测,得到48种预测模型,根据各模型的平均判定系数对模型进行初步筛选;最后,综合比较归一化均方误差以及平均绝对百分比误差,得到2组待调优模型,分别对其进行超参数调优,得到最优预测模型.结果表明:最优预测模型的判定系数为0.971,归一化均方误差为0.029,平均绝对百分比误差为5.3%.To improve the prediction accuracy of coal mine safety situations,a coal mine safety situation prediction model based on the fusion of feature selection and machine learning is proposed.Firstly,the relevant data of China's coal mine safety production situation from 1978 to 2019 are normalized,and the 8 types of influencing factors of coal mine safety production situation are merged by using full subset regression and random forest feature selection methods.Then,12 combinations of distinctive characteristics of the coal mine safety production condition are predicted by using 4 machine learning methods,including ridge regression,classification and regression tree,AdaBoost,and XGBoost,then 48 prediction models are obtained.The models are preliminarily screened according to the average judgment coefficient of each model.Finally,by comprehensively comparing the normalized mean square error and the average absolute percentage error,two optimized models are obtained,and their hyperparameters are optimized respectively to obtain the optimal prediction model.Results show that the R~2 of the best optimal prediction model is 0.971,the normalized mean square error is 0.029,and the average absolute percentage error is 5.3%.

关 键 词：煤矿安全生产态势 特征选取 机器学习 预测 

分 类 号：X936[环境科学与工程—安全科学]