本文以煤层顶板冒落数据作为样本,利用weka等软件,利用粗糙集理论等相关非线性理论知识,分别采用支持向量机和Bp-神经网络对煤层顶板冒落进行训练和预测。从详细的精度,混淆矩阵和节点错误率这三个方面分别比较三种算法,从而得到结论是:Bp-神经网络的训练和预测效果是稍微强于支持向量机。 In this paper, taking the coal seam roof falling data as a sample, this paper uses the software of weka and other relevant nonlinear theoretical knowledge of rough set theory to train and predict coal seam roof caving using support vector machine and Bp-neural network respectively. Comparing the three algorithms respectively from three aspects: detailed accuracy, confusion matrix and node error rate, we conclude that the training and prediction effects of Bp-NN are slightly better than SVM.