Fraud in the healthcare industry remains one of the most profound issues globally, with billions lost each year. This research focuses on one of the most recent methods, applying Machine Learning (ML) algorithms, to automate the detection of fraudulent claims. Multiple models including Logistic Regression, Random Forest, and XGBoost and Neural Networks as well as Isolation Forest were trained and evaluated on an accuracy dataset that contained 1 million claims insurance. Several challenges such as redundant features and noisy data were successfully overcome using data preprocessing and visualization methods. These models were compared on the basis of accuracy, precision, recall, and F1-score. XGBoost achieved the greatest accuracy of 92.1%, and Neural Networks and Random Forests followed close behind. Although unsupervised, Isolation Forest proved useful in identifying anomalous patterns in situations with little labeled data. ML techniques require model interpretability and computational efficiency to detect fraud in real-world healthcare systems.