Medical insurance fraud is a serious challenge in the healthcare industry, and its rapid escalation necessitates improved detection mechanisms. This study examines three machine learning techniques-supervised (Random Forest), unsupervised (K-means clustering), and hybrid learning-for detecting medical insurance fraud across two datasets from some National Health Insurance Scheme (NHIS)-approved hospitals in Ghana concerning insurance claims. The performance indicators for these strategies are as follows: Random Forest attained detection accuracies of 91% and 93%, K-means clustering generated 70% and 46%, and the hybrid model produced 34% and 42%, respectively. The growing prevalence of medical insurance fraud highlights the critical need for effective detection methods to protect the integrity of healthcare systems.