基于数据填补-机器学习的煤与瓦斯突出预测效果研究*

为解决煤与瓦斯突出事故数据集少，数据缺失严重的问题，提出将多重插补（MI）和随机森林填补(MF)应用于填补缺失参数，并将填补前和填补后的数据输入SVM，ELM，RF 3种机器学习算法进行训练，构建9种耦合模型。采用总体准确率、局部准确率、运行时间这3种指标评价模型性能。研究结果表明:采用数据填补算法后，由于训练样本增大，煤与瓦斯突出事故预测的总体准确率提高，运行时间增长；MF-RF模型的总体准确率与事故预测准确率最高，分别为97.90%和98.93%；RD-ELM模型的运行时间最短，为0.24 s；多重插补使得煤与瓦斯突出预测的总体准确率提高0.98%~1.11%，随机森林填补总体准确率提高5.13%~7.50%，随机森林填补的效果好于多重插补。

Study on prediction effect of coal and gas outburst based on data imputation and machine learning

In order to solve the problem of less data sets and serious data missing in the coal and gas outburst accidents,the multiple imputation (MI) and MissForest (MF) were applied to fill the missing parameters.The data before and after imputation were input into three machine learning algorithms of SVM,ELM and RF for training,then nine coupling models were constructed,and the overall accuracy,local accuracy and running time were used to evaluate the performance of the models.The results showed that after using the data imputation algorithms,the overall accuracy of coal and gas outburst accident prediction improved,and the running time increased due to the increase of training samples.The overall accuracy and accident prediction accuracy of MF-RF model were the highest,which was 97.90% and 98.93%,respectively.The running time of RD-ELM model was the shortest as 0.24 s.The MI improved the overall accuracy of coal and gas outburst prediction by 0.98%~1.11%,and the overall accuracy of MF increases by 5.13%~7.50%.The effect of MF was better than that of MI.