Machine Learning Algorithms with Oversampling Methods for Automobile Insurance Fraud Detection

Some countries regulate purchasing vehicle insurance coverage as mandatory to enable risk transfer and financial protection. With the ever-expanding business-related scope of auto insurance, the crime in fraud insurance is a vital task to fight. The imbalanced dataset makes predicting fraud on automobile insurance claims challenging. Most fraud detection techniques result in inefficient classification models with an imbalance issue. This work is concerned with developing a fraud detection system to examine the increase in model performance with the help of the Synthetic Minority Oversampling Technique (SMOTE) and Adaptive Synthetic Sampling (ADASYN) to remove the imbalance. The abilities to predict series machine learning models were compared by tenfold cross-validation, namely Support Vector Machine (SVM), Random Forest (RF), Balanced Random Forest (BRF), and Extreme Gradient Boosting (XGBoost) Classifier. The majority voting approach is employed to ensemble numerous classifiers to increase prediction performance in the final stage. As a result, all models achieve satisfactory results, and an F1score of more than 98% has been achieved using Support Vector Machine (SVM) model, which employs the SMOTE oversampling method.