基于PCA-PNN的采空区多源指标危险性辨识*

为了提高采空区多源指标危险性辨识的预测精度，基于主成分分析（PCA）和概率神经网络（PNN），提出1种采空区多源指标危险性辨识方法。将影响华东某地区矿山采空区危险性辨识的9项因素作为主要影响因素，并以96个实测采空区为例进行分级。研究结果表明:与朴素贝叶斯、随机森林和AdaBoost 3种机器学习算法相比，PNN在测试集上表现更好，对实际工程具有良好的指导意义和应用价值。

Multi-source indexes risk identification of goaf based on PCA-PN

In order to improve the prediction accuracy for the multi-source index risk identification of goaf,based on the principal component analysis (PCA) and the probabilistic neural network (PNN),a kind of risk identification method of the multi-source indexes of goaf was proposed.9 factors affecting the risk identification of goaf in a region of east China were determined as the primary influencing factors,and 96 measured goafs were classified as examples.The results showed that compared with three machine learning algorithms of the Naive Bayes,the Random Forest and the AdaBoost,PNN performed more preferable on the test set,which has admirable guiding significance and application value for the practical engineering.