Optimizing Heart Disease Prediction: ESAPSO-BP Neural Network Algorithm Integration for Enhanced Accuracy and Robustness

Abstract

Heart disease is a major problem plaguing human beings, especially the middle-aged and elderly people. In the world, the incidence of heart disease is high. In the era of artificial intelligence, through machine learning algorithms, people can build models according to various modal characteristics of patients to predict the disease, so as to assist doctors to make judgments about the condition of patients. Aiming at the problem that the accuracy of heart disease prediction model is not high, this paper proposes an ESAPSO-BP neural network algorithm combining particle swarm optimization algorithm and simulated annealing algorithm, and introducing particle elimination strategy. In order to compensate for the lack of feature selection,we use One-hot coding and PCA technology to select features. After the verification of two data sets, UCI and Kaggle, the accuracy of our proposed method on the two data sets reaches 93.22% and 95.20% respectively, which has high accuracy and robustness, especially in small samples. The model can provide help for doctors and patients to predict and treat heart disease.