Prediction Of Chronic Liver Disease Based On Principal Component Analysis Machine Learning

Chronic liver disease is one of the most dangerous infectious diseases,which is prevalent with a low cure rate and a high mortality rate.For newly admitted patients with chronic liver disease,timely and accurate assessment of the disease and the development of treatment programs have become a pressing issue to be studied and resolved.In this paper,according to the patient's test indicators,analysis and mining to obtain a better prediction model for the doctor to provide auxiliary diagnosis.It is an effective way to improve the diagnosis rate of chronic liver disease by using a new generation of information technology to predict the mechanism and characteristics of chronic liver disease.Based on the existing disease prediction,this paper first uses Receiver Operating Characteristic(ROC curve)and Logistic regression analysis to judge the value of diagnosis of which test indicator of chronic liver disease in the light of the advantages and limitations of the existing prediction methods;and then uses the principal component analysis machine learning algorithm to do the linear term dimension reduction process for the description of the property item number of indicators of chronic liver disease;Finally,the ant colony algorithm is used to improve the BP neural network training learning to obtain the prediction model.In this paper,a series of pre-processing design of the original data is presented,and a prediction model of chronic liver disease based on optimal indicators and principal component analysis is proposed.The results of the two groups are analyzed by using the ROC curve and the Logistic regression method.The 13-dimensional test indicators are reduced to 5-dimensional data by principal component analysis(PCA);the ant colony algorithm is used to improve the BP neural network training and get the optimal model.Based on the pre-processing design of the original indicators data set,three different input prediction models are obtained,which are:A.20-dimensional primitive indicators items;B.The 13-dimension test indicators items based on ROC curve and Logistic regression analysis;C.5-dimension comprehensive data set based on ROC curve,logistic regression analysis and principal component analysis.Taking these three data sets as the input of neural network for comparative analysis,the following conclusions are obtained:(1)The prediction accuracy of the C is the highest in these three models,which is 15.2%higher than that of the A,which is 9.3%higher than that of the B;(2)Based on ROC curve and Logistic regression analysis,comprehensive evaluation of the diagnostic value of chronic liver disease diagnosis value,help to reduce information redundancy and interference;(3)The neural network model based on principal component analysis reduces the prediction complexity and the topology of BP neural network input;(4)Optimize the network structure and ant colony algorithm to improve the BP neural network,greatly improving the convergence rate and the prediction accuracy.