.png){kind=icon}
.png){kind=icon}
Return
Enhancing heart disease prediction using genetic algorithm-based ensemble learning with explainable AI
.png){kind=icon}
PRE
AI
DOI：10.1088/2631-8695/ae6f88.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Heart disease is one of the major causes of deaths in the world. Although the existing techniques like genetic algorithm-support vector machine (GA-SVM) are very accurate, they are not easy to interpret and can not be used across different datasets. This paper suggests a new GA-explainable artificial intelligence (XAI) Ensemble Framework consisting of the combination of feature selection by GA and a heterogeneous ensemble of SVM, random forest, and extreme gradient boosting. The obtained final decision is obtained by a stacking meta-classifier, in order to minimize variance and bias. XAI algorithms (Shapley additive explanations and local interpretable model-agnostic explanation) were also added to gain quality clinical usability, which gives insight into the role of features. The model was tested on the Kaggle cardiovascular disease dataset which consists of more than 70 000 patient records. The experimental results showed that the proposed GA-XAI Ensemble scored at 99.1% accuracy, 98.7% F1-score, and area under the curve of 0.991, which were better than the existing GA-SVM based approaches. Moreover, type of chest pain as well as thalach and serum cholesterol were the most significant characteristics emphasized in the analysis of interpretability. The method provides a better predictive power and clinically meaningful information, and should be used as it is appropriate to implement the method in practice as a part of a health care decision support system.
Keywords:
heart disease prediction
genetic algorithm (GA)
explainable artificial intelligence (XAI)
random forest (RF)
extreme gradient boosting (XGBoost)
stacking meta-classifier
clinical decision support system