Using AI Algorithm to Establish the CVD Risk Assessment Model

In Taiwan, diseases with cardiovascular include heart disease and cerebrovascular disease among the top ten causes of death. With the development of data mining in the medical field, it can be used to establish the risk prediction models of disease as a tool to assist physicians in decision-making in diagnosis. The purpose of this study is taking Taiwanese health examination data as an example, to identify the significant risk factors of cardiovascular disease and to establish the risk assessment model of cardiovascular disease using data mining. Using Chi-square test and Information Gain identify the correlation between various factors and cardiovascular diseases. Using eight algorithms such as decision tree, random forest, XGBoost, neural network, logistic regression, support vector machine, K nearest neighbor algorithm and voting algorithm to establish the risk assessment model of cardiovascular disease and using confusion matrix and AUC as model evaluation. Compare model performance with different factor combinations. Experiment result shows that it finds 22 questionnaires and biochemical variable risk factors affecting cardiovascular disease and 10 questionnaire factors. ANN and VOTE with 22 factors of Information Gain (threshold > 0.01) are the best models. Both models have the same accuracy (0.88) and AUC (0.90). The best model of questionnaire variable is ANN with the accuracy rate is 0.89, and the AUC is 0.91.