Abstract: Insurance claim fraud detection is a serious problem, resulting in heavy financial loss to insurers and impacting policyholder premium levels. The current methods for fraud detection are rule-based and manual inspection-based and are mostly inefficient and error-prone. The use of Machine Learning (ML) methods to enhance the detection and investigation of fraudulent insurance claims is explored in this paper. Employing several ML algorithms, i.e., supervised learning techniques including Random Forests, Support Vector Machines, and neural networks, we illustrate how fraud activity from historical claims data can be discovered and leveraged to predict fraud risk. The article describes the preprocessing of the claims data, feature engineering, model estimation, and validation procedures in the design of successful fraud detection models.

| DOI: 10.17148/IJIREEICE.2025.13319