Abstract
Currently, there are over 300 million patients with cardiovascular diseases in China. With the acceleration of population aging, the impact of cardiovascular diseases is becoming increasingly severe. Accurately and efficiently predicting the potential risks of cardiovascular disease is crucial for preventing its progression and maintaining public cardiovascular health. This article uses a combination of gradient-boosting decision trees (GBDT) and logistic regression (LR) to predict the probability of cardiovascular disease risk. To address the weak feature combination ability of LR in handling nonlinear data, a cardiovascular disease prediction model was established by integrating GBDT with LR by using the predicted results of GBDT as new features instead of the original ones and inputting them into the LR model. Using the UCI cardiovascular disease dataset, we conduct experimental comparisons between the proposed model and other common disease classification algorithms such as logistic regression (LR), random forest (RF), and support vector machine (SVM). The experimental results show that GBDT+LR outperforms other models in multiple evaluation indicators such as accuracy, precision, specificity, F1 value, MCC, AUC, and AUPR. The cardiovascular disease prediction model using the GBDT+LR algorithm has the best prediction performance. This article builds a front-end and back-end separated cardiovascular disease analysis and prediction platform based on the Spark Big data framework and Vue+SpringBoot framework, which realizes predicting cardiovascular disease risk probability.