Background: Delayed or misdiagnosis of myocardial infarction (MI) is a common occurrence in clinical settings. Timely detection of MI is crucial to prompt intervention which minimizes irreversible heart muscle damage, thereby reducing the risk of complications and heart failure. Thus, our objective was to develop a machine learning (ML) model to serve as a diagnostic aid, utilizing a minimal set of patient health parameters as features. Methods: We collected data from 1,200 individuals (300 MI, 900 non-MI) using a case-control study with a 1:3 case-to-control ratio. Employing three feature selection methods, we identified significant variables. Six ML models (Naïve Bayes, Logistic Regression, Decision Tree, SVM, Random Forest, AdaBoost) were constructed for each technique, and their performance was evaluated using F1-Score, Cohen's Kappa, and AUROC. Additionally, clinical validation was conducted on real-time data for practical applicability. Results: 17, 18, and 9 features were selected using variance threshold, correlation-based, and a combination of both techniques respectively. AdaBoost consistently showcased superior performance, followed by Random Forest. In real-time clinical validation, AdaBoost demonstrated remarkable performance with 94.12% accuracy, 98.86% precision, 98.58% recall, 93.11% F1 score, 96.49% Cohen's Kappa, and 94.12% Area under ROC. Conclusion: The ML can serve in a timely, and precise diagnosis of MI, particularly AdaBoost. Furthermore, identified risk factors and their correlations emphasize the need for personalized preventive actions and lifestyle changes to mitigate myocardial infarction risks.