Building Robust Fraud Detection Model for Insurance Claims Using SMOTE

The insurance fraud claim detection is a critical task in the insurance industry. Various methodologies were proposed in terms of detecting fraud claims but the main thing is handling the proportionality between the fraud and non-fraud cases in the dataset before it was given to the model, in that case class imbalance may occur. Class imbalance where the number of fraudulent claims is significantly lower than legitimate claims which pose a challenge for accurate fraud detection. In this paper, we employed [Elreedy and Atiya in Inf Sci 505:32–64, 2019] Synthetic Minority Over-sampling Technique (SMOTE) algorithm to address the imbalance problem and enhance the performance of the fraud claim detection model. SMOTE is a popular technique for oversampling the minority class by creating synthetic samples that are similar to the existing minority class instances. By generating synthetic examples, SMOTE helps in balancing the class distribution and allows the model to learn from a more representative dataset. This technique is particularly effective when the available data is limited and insufficient to capture the complexities of the limited class. This process results in a larger and more balanced dataset, enabling the model to learn from a diverse range of fraudulent claim patterns. By applying SMOTE to our dataset, we are able to overcome the class imbalance issue and improve the performance of the fraud claim detection model. The resampled dataset provides a more accurate representation of the underlying distribution, leading to enhanced detection of fraudulent claims. We evaluate the performance of the model by measuring various metrics such as accuracy, precision, recall and F1-score. Our findings demonstrate the effectiveness of SMOTE in addressing class imbalance with improved random forest and LightGBM model for fraud claim detection process. The utilization of SMOTE contributes to better identification of fraudulent insurance claims, reducing potential losses and enhancing the overall efficiency of the insurance industry.