A Joint Method Of Sparse Principal Component Analysis And Back-propagation Neural Network With Its Application To Heart Disease Clinical Data Analysis

Heart Disease is widely considered as the first most common cause of mortality over the world.More than 4 million people die of heart disease every year in China.The adverse consequences of heart have led to global interest and work for improving the management and diagnosis of heart disease.Various techniques for statistical methods and models have been used for accurate prediction of occurrence of heart disease based on the risk factors that are associated with the clinical records of the heart disease patients.However,the prediction speed and accuracy of existing heart disease prediction models did not meet our expectations.According to the research results in recent years,the hybrid algorithm has shown good performance in the prediction of heart disease,and the prediction speed and accuracy have been significantly improved.Therefore,we propose a new algorithm from the perspective of hybrid models to improve the model prediction speed and accuracy.First,we use sparse principal component analysis to reduce the dimension of the data.Then,the sparse principal component score is used as the input value of the back-propagation neural network,which is then trained by the back-propagation neural network.Combined with the back propagation neural network,we propose a new algorithm,named SPCA-BP algorithm.We evaluated our proposed method in a heart disease clinical data from the Cleveland Clinic.Considering the probabilistic characteristics of disease prediction problems,we completed the transformation from logistic regression to neural network,and completed the analysis combined with our method.At the same time,we also applied our method to the COX proportional hazards model,and calculated the accuracy of the model prediction.As a comparison,we also performed logistic regression analysis on the data.The results show that our algorithm works best in predicting diagnosis.