Meta-Learning Based Classification Model for Cardiovascular Disease (Preprint)

BACKGROUND Cardiovascular disease is a significant global health concern, being the leading cause of death and disability worldwide. The World Health Organization reports that cardiovascular disease accounts for 17.9 million deaths each year, representing 31% of all global deaths. Heart disease, in particular, is a major contributor to mortality worldwide. Machine learning algorithms have shown promise in predicting the risk of heart attacks. One particular method, meta-learning, is a type of machine learning that enables a system to learn how to learn. Meta-learning encompasses a set of techniques that allow a system to improve its own learning process. OBJECTIVE In this paper, our objective is to propose a meta-learning-based classification model for cardiovascular diseases, specifically for heart attack classification. We aim to utilize a dataset containing 76 attributes, with the predicted attribute being the presence of heart disease. METHODS To achieve our objective, we follow the following steps: We gather a dataset with 76 attributes, including information related to cardiovascular health and the presence of heart disease. We evaluate traditional classification models commonly used in heart attack classification. We implement a meta-learning approach to enhance the accuracy of heart attack prediction. We compare the results obtained using the meta-learning approach with the traditional classification models. Additionally, we explore the impact of using Synthetic Minority Over-sampling Technique (SMOTE) to balance the target classes in the dataset and compare the results with and without SMOTE. RESULTS Our results demonstrate that the meta-learning approach outperforms traditional classification models in predicting heart attack risk. The accuracy of the meta-learning model is significantly higher compared to the traditional models we evaluated. Furthermore, we observe that using SMOTE to balance the target classes improves the performance of the meta-learning model even further. CONCLUSIONS Based on our findings, we conclude that the meta-learning approach is highly effective for heart attack classification. The use of meta-learning techniques enhances the accuracy of heart attack risk prediction compared to traditional models. Furthermore, incorporating SMOTE to balance the target classes improves the overall performance of the meta-learning model. These results suggest that the meta-learning approach can be leveraged to improve the accuracy and effectiveness of cardiovascular disease prediction and classification models, specifically for heart attack risk assessment.