The Rearch on Improved LVQ Neural Network Method

Abstract

By analyzing the basic structure and algorithm of LVQ neural network and its deformation, this paper studies the breast tumor diagnosis method based on improved LVQ neural network. The LVQ neural network needs to process all input features, because the input feature dimension is too large, the redundant information is too much, the training takes too long, so apply PCA algorithm to preprocess the input features. In essence, PCA algorithm takes the direction with the largest variance as its main feature and “de-correlates” the data in each orthogonal direction. It is a parameterless technology without the intervention of subjective parameters, so PCA is convenient for general implementation. In this paper, PCA algorithm is proposed to optimize LVQ neural network, and the results are compared with LVQ neural network. The results show that the optimized LVQ neural network by PCA has shorter running time and higher diagnostic accuracy than LVQ neural network.