Open Access

ARTICLE

Risk Indicator Identification for Coronary Heart Disease via Multi-Angle Integrated Measurements and Sequential Backward Selection

Hui Qi¹, Jingyi Lian², Congjun Rao^2,*

1 School of Information Engineering, Sanming University, Sanming, 365004, China
2 School of Mathematics and Statistics, Wuhan University of Technology, Wuhan, 430070, China

* Corresponding Author: Congjun Rao. Email:

Computer Modeling in Engineering & Sciences 2025, 145(1), 995-1028. https://doi.org/10.32604/cmes.2025.069722

Received 29 June 2025; Accepted 12 September 2025; Issue published 30 October 2025

Abstract

For the past few years, the prevalence of cardiovascular disease has been showing a year-on-year increase, with a death rate of 2/5. Coronary heart disease (CHD) rates have increased 41% since 1990, which is the number one disease endangering human health in the world today. The risk indicators of CHD are complicated, so selecting effective methods to screen the risk characteristics can make the risk prediction more efficient. In this paper, we present a comprehensive analysis of CHD risk indicators from both data and algorithmic levels, propose a method for CHD risk indicator identification based on multi-angle integrated measurements and Sequential Backward Selection (SBS), and then build a risk prediction model. In the multi-angle integrated measurements stage, mRMR (Maximum Relevance Minimum Redundancy) is selected from the angle of feature correlation and redundancy of the dataset itself, SHAP-RF (SHapley Additive exPlanations-Random Forest) is selected from the angle of interpretation of each feature to the results, and ARFS-RF (Algorithmic Randomness Feature Selection Random Forest) is selected from the angle of statistical interpretation of classification algorithm to measure the degree of feature importance. In the SBS stage, the features with low scores are deleted successively, and the accuracy of LightGBM (Light Gradient Boosting Machine) model is used as the evaluation index to select the final feature subset. This new risk assessment method is used to identify important factors affecting CHD, and the CHD dataset from the Kaggle website is used as the study subject. Finally, 11 features are retained to construct a risk assessment indicator system for CHD. Using the LightGBM classifier as the core evaluation metric, our method achieved an accuracy of 0.8656 on the Kaggle CHD dataset (4238 samples, 16 initial features), outperforming individual feature selection methods (mRMR, SHAP-RF, ARFS-RF) in both accuracy and feature reduction. This demonstrates the novelty and effectiveness of our multi-angle integrated measurement approach combined with SBS in building a concise yet highly predictive CHD risk model.

---

Keywords

---