摘要
逐步线性回归能较好地克服多重共线性现象的发生,因此逐步回归分析是探索多变量关系的最常用的分析方法,智能算法是现代数据分析的主要方法。本文通过一个实例进行了对比研究,预测结果显示:在预测的精度上,在隐含层数目相同时,RBF径向神经网络>BP神经网络>逐步线性回归>ELM极限学习机。通过对比分析,发现神经网络方法较回归分析预测效果更好,误差相对较小。
Gradient linear regression can well solve the occurrence of Muhicollinearity, so the gradient regression analysis is analytical method to research the correlation among muhivariable. Intelligent algorithm is one of the dominant methods in modern data analysis. Both of the methods above are applied to one example and further to be compared. The forecasted result shows ： for the accuracy of the forecasted results, when the number of hidden layer is consistent, RBF radial basis neural networks 〉 BP neural networks 〉 Gradient linear regression 〉 ELM limit machine learning. Through the analysis of comparison, we infer that the accuracy and error of neural networks is smaller than the regression model.
出处
《华北科技学院学报》
2014年第5期60-65,共6页
Journal of North China Institute of Science and Technology
基金
华北科技学院<煤矿安全评价数学方法研究(创新团队)>(3142014127)
华北科技学院重点学科应用数学(HKXJZD201402)
