Research On Classification Algorithm In Heart Disease Predicting And Diagnosing

In recent years,the rapid development of technologies such as big data and machine learning has brought innovation and change to many industries.With the development of precision medical plans,the combination of machine learning and health care big data has brought people the infinite imagination of the future big data health business.In order to reduce the risk that doctors may misdiagnosis due to their lack of experience,the classification algorithm in machine learning can be applied to the auxiliary diagnosis of the disease.Using the results of model classification to help doctors make judgments can improve the accuracy of doctors' diagnosis.The classification algorithm first trains the corresponding classification model for the training set,and then inputs the patient's examination data into the trained classification model to judge whether the patient has such diseases.Although the classified pre-diagnosis model cannot completely replace the doctor,doctors can make more accurate judgments by referring to the classification results.Heart disease is the leading killer of human health,and one third of the world's population is caused by heart disease.In China,hundreds of thousands of people die of heart disease every year.Therefore,if you can use the relevant measurable indicators of the human body and use big data research methods to predict the possibility of heart disease,it will play a vital role in understanding people's health and preventing heart disease.The purpose of this paper is to find a classification model that is relatively effective for pre-diagnosis of heart disease by comparing the prediction accuracy of several different classification algorithms on the heart disease datasets and its parameters.The main research contents and achievements are as follows:1.By comparing the classification accuracy of K-nearest neighbor,linear kernel SVM,RBF kernel SVM,logistic regression,decision tree,naive Bayes and random forest on the heart disease datasets.We find that RBF kernel SVM has a higher classification accuracy in the heart disease pre-diagnosis.The conclusion will serve as the basis for subsequent studies.2.Considering that the classification accuracy rate of RBF kernel SVM is greatly affected by its parameter combination.The improved APSO algorithm is used to replace the grid search method to optimize its parameters.An improved RBF kernel SVM heart disease pre-diagnosis model based on APSO algorithm is proposed.3.The improved RBF kernel SVM classification model based on APSO algorithm was applied to the classification of heart disease datasets.The results confirmed that the classification accuracy was further improved.The classification accuracy rate increased from 85.56% to 86.55% in the Cleveland Clinic's heart disease dataset,and from 87.78% to 88.80% in the Hungarian Institute of Cardiology's heart disease dataset.The classification pre-diagnosis models for heart disease datasets presented in this paper have been verified on the UCI datasets respectively.