An Efficient Predictive Model for Myocardial Infarction Using Cost-sensitive J48 Model

Area of Science:

• Medical Informatics
• Data Mining
• Cardiology

Background:

• Myocardial Infarction (MI) is a life-threatening condition with significant human and economic costs.
• Predicting MI is crucial for timely intervention and improved patient outcomes.
• Existing prediction models often struggle with the inherent imbalance in patient data.

Purpose of the Study:

• To develop and evaluate a novel data mining model for predicting Myocardial Infarction.
• To address the challenge of imbalanced datasets in MI prediction.
• To enhance the accuracy and reliability of MI prediction models.

Main Methods:

• A hybrid feature selection method combining Relief and Genetic Algorithm was employed.
• A cost-sensitive classifier (Metacost with J48) was utilized to handle misclassification costs.
• The model was trained and tested on a dataset of 750 patients (455 healthy, 295 MI cases).

Main Results:

• The optimal cost ratio of 1:200 yielded the best performance compared to models without feature selection or cost-sensitivity.
• The developed model achieved a sensitivity of 86.67%, F-measure of 80%, and accuracy of 82.67%.
• The integration of cost-sensitive learning and feature selection demonstrably improved prediction outcomes.

Conclusions:

• The proposed cost-sensitive prediction model, enhanced by hybrid feature selection, proves effective for Myocardial Infarction.
• This approach offers a reliable tool for predicting MI, particularly in imbalanced datasets.
• The findings support the clinical utility of advanced data mining techniques in cardiovascular disease prediction.