论文部分内容阅读
为了提高煤与瓦斯突出预测的准确度,提出了一种基于方差比检验的预测器输入主因素识别方法.在一定显著水平下对增添或删除若干因素前后预测器的预测残差进行F检验,用以确定具有改进作用的增添或删除操作.遍历增添和删除的所有情形后,即可确定能获得最大改进的主因素输入组合.所提出方法对任何预测器都适用,以一个基于人工神经网络的煤与瓦斯突出预测器为例,结合两个煤矿的煤与瓦斯突出影响因素的实测样本,采用所建议的方法对进行了输入因素遴选,结果表明:采用得到的主因素作为预测器的输入比采用全部因素作为输入因素得到的预测结果更加准确,表明所建议的主因素识别方法是可行的,并且有助于改进预测器的精度.
In order to improve the accuracy of prediction of coal and gas outburst, a predictor input principal factor identification method based on variance ratio test is proposed.F-test is performed on the prediction residuals of the predictor before and after adding or deleting a certain number of factors, To determine which additions or deletions have the effect of improvement After traversing all cases of additions and deletions, you can identify the input combination of the main factors that will yield the most improvement The proposed method applies to any predictor, based on an artificial neural network Taking the coal and gas outburst predictor as an example, we select the input factors by using the proposed method based on the measured samples of coal and gas outburst in two coal mines. The results show that the main factor is used as the input of the predictor This result is more accurate than using all the factors as input factors, which shows that the proposed method of identifying principal factors is feasible and helps to improve the accuracy of the predictor.