Liver Disease Based On Principal Component Analysis And Neural Network Prediction Analysis

In this paper, the model based on principal component analysis (PCA) and neural network prediction is proposed on the ground that analyzing the existing liver forecasting method and against the limitations of the existing prediction method. In order to reduce the complexity of forecasting, the first we should reduce the order of the diagnosis targets, and then build the prediction models based on BP neural network and RBF neural network. The main research in this paper is as follows:Due to the need for objective diagnostic data has 13 items, which will undoubtedly increase the complexity of BP network prediction model. On the premise of retaining the original information we should take seven parameters instead of 13 parameters using the principal component analysis. It avoid that the network model is too complicated to impact the result of prediction.This paper introduces the basic principles of BP neural network. For the problems of BP network in forecasting, we proposed effective solution separately. First we should take initial value normalization processing, and then select the best hidden nodes using trial and error method according to commonly used selected formula. For the problem that slow velocity of convergence, we use L-M algorithm instead of gradient descent training of neural networks. Also we have improved prediction accuracy based on the BP neural network experiments, and used a local approximation of the RBF neural network to further improve prediction.Based on principal component analysis and neural networks, we have established a liver disease prediction model, and achieved good prediction results, and had laid a good foundation in order to further improve predictive power, further study of the issue.