A Novel Hybrid Method for Imbalanced Automobile Insurance Fraud Detection
Phannana Aiemsuwan, Supawadee Srikamdee · 2024
This paper addresses fraud detection in the automobile insurance sector, which experiences the highest rate of fraudulent claims in the industry. The study proposes a methodology that combines resampling techniques and backward elimination for feature selection to tackle data imbalance. This approach significantly improves the accuracy of machine learning models in identifying fraudulent activities. The research conducts a comprehensive comparison of seven major machine learning algorithms: Logistic Regression, Decision Tree, Random Forest, k-nearest Neighbors, Naive Bayes, XGBoost, and Support Vector Machine. These models are evaluated based on precision, recall, and F1 score across three different datasets to determine their effectiveness in fraud detection. The findings reveal that the Random Forest algorithm is the most efficient, consistently achieving F1 scores above 0.92. This highlights the vital role of machine learning in detecting and preventing fraud in the automotive insurance industry.