基于主成分分析的GA-BP神经网络地表下沉系数预测

Prediction of Surface Subsidence Coefficient by Improved BP Neural Network Based on Principal Component Analysis

为提高条带开采地表下沉系数预测准确率,基于地表下沉系数影响因素具有一定相关性、不确定性以及非线性的复杂现象,建立基于主成分分析的遗传算法(Genetic Algorithm,GA)优化的神经网络GA-BP智能预测模型。利用遗传算法对BP神经网络(Back Propagation Neural Network)的初始权值和阈值进行优化处理,通过SPSS20(Statistical Product and Service Solutions 20)软件对地表下沉系数影响因素进行主成分分析,降低数据维度,消除变量间的冗余信息,找出主成分并作为模型的输入样本,利用MATLAB(Matrix Laboratory)软件进行仿真与分析。结果表明:与传统BP神经网络模型和主成分分析(Principal Component Analysis,PCA)的PCA-BP神经网络模型相比,基于主成分分析的GA-BP模型的相对误差不超过5%,与实测值更为接近,预测精度进一步提高,基本满足矿区实际工程需要,为条带开采地表下沉系数预测提供了又一种准确可行的方法。

In order to improve the accuracy of prediction of surface subsidence coefficient in strip mining,an intelligent prediction model of GA-BP based on principal component analysis was established,which was based on the complex phenomena of correlation,uncertainty and non-linearity of influencing factors of surface subsidence coefficient.The genetic algorithm was explored to optimize the initial weights and thresholds of BP neural network.The principal component analysis of influencing factors of surface subsidence coefficient was carried out by SPSS20 software,which reduced the data dimension,eliminates redundant information among variables,finds principal components and takes them as input samples of the model.The simulation and analysis were carried out by using MATLAB software.The results show that compared with the traditional BP neural network model and PCA-BP model,the relative error of GA-BP model based on principal component analysis is less than 5%,which was more close to the measured value,and the prediction accuracy was further improved,which basically met the actual engineering needs of mining area,and provided another accurate and feasible method for prediction of surface subsidence coefficient in strip mining.