Predicting Heart Disease Using Automated Machine Learning Based on Genetic Algorithms

This study aims to apply automatic machine-learning approaches using genetic algorithms to enhance heart disease prediction. Heart disease has remained the major cause of mortality in the world, necessitating an effective and timely diagnosis. Most current diagnostic and assessment processes are lengthy and expensive, relying heavily on clinical expert knowledge. To help address these issues, machine learning approaches, which derive their utility from examining substantial datasets for the recognition of patterns, have emerged as a potential solution, providing solutions beyond those achievable by human recognition alone. Genetic algorithms are also suited to addressing these issues as they mimic natural evolution to perfect high-caliber machine-learning models, feature selection, and parameter selection in machine-learning applications. This study examines the utilization of genetic algorithms working alongside AutoML frameworks to improve accuracy in heart disease predictions. Reducing to the best combination of attributes and the optimum parameters for each attribute is a time-consuming task, so automating this aspect of the process allows for more accurate and prompt predictions, consequently reducing the manual work. The AutoML approach followed in this research is TPOT, which uses genetic algorithms to ascertain optimally designed machine-learning pipelines. The application of AutoML, together with genetic algorithms, is the most prominent finding that yields a significant improvement in the quality of the predictions for heart disease compared to the traditional assessment approaches, with an accuracy of 93.8%. This approach will enhance diagnostic accuracy and enable early diagnosis, thereby reducing the likelihood of misdiagnoses or ineffective treatments and ultimately lowering associated costs.