A Comprehensive Review of Machine Learning Approaches for Predictive Analytics in Healthcare Diagnosis and Clinical Decision-making
Résumé
This study explores the application of machine learning (ML) methodologies to improve clinical decision-making, with a specific emphasis on diagnostic and prognostic modeling in the healthcare domain. It underscores the utility of ML algorithms in processing large-scale, heterogeneous medical datasets for disease classification, risk prediction, and pattern recognition. Despite their potential, several challenges hinder the full integration of ML in clinical settings, including model interpretability, class imbalance (particularly in datasets with low disease prevalence), and the necessity of earning clinicians’ trust through reliable and transparent predictions. The research implements traditional ML techniques to support prognostic assessment in life-threatening cardiovascular conditions, addressing key preprocessing tasks such as feature engineering, class rebalancing, and threshold optimization for risk stratification. Furthermore, the study incorporates a patient-centered framework by modeling patient preferences using supervised learning algorithms to predict individualized treatment choices. This dual approach—combining technological precision with human-centric considerations—aims to assist clinicians in delivering personalized care, thereby enhancing decision quality, patient engagement, and overall satisfaction with healthcare outcomes.