Variables Screening Methods Based on the Optimization of RBF Neural Network
基于优化的RBF 神经网络的变量筛选方法

Because of the Characteristics of RBF neural network structure is simple, output and initialized weights irrelevant, adaptive, less adjustable parameter etc. This paper proposes using the method of cross validation to find the optimal parameter value of SPREAD, constructs the optimal RBF neural network model and combines the algorithm of MIV to use for variables screening. Through the example test the validity of the model, also make the method has better ability of stability and applied.