Fraud Detection in Auto Insurance Claims using Bootstrap Aggregating Classifier

TRESSYA SHAFA AMARA AHADI, Drs. Danardono, MPH., Ph.D.

2024 | Skripsi | S1 ILMU AKTUARIA

Insurance fraud is an essential and costly challenge for all insurance-related industries. In Indonesia, insurance claim fraud is still widespread. The number of fraudulent insurance claims has made insurance companies focus on minimizing fraud. Nevertheless, insurance companies find it challenging to manually verify each claim due to the high number of claims. Therefore, this study will use machine learning techniques to identify fraud in an auto insurance claim. In this study, fraud detection is carried out using the Bootstrap Aggregating classifier method with three base learners, namely Logistic Regression, Support Vector Machines, and Naïve Bayes—implementation of Bootstrap Aggregating on the base learner to improve model performance without overfitting. Based on the analysis conducted, it is concluded that Bootstrap Aggregating on Support Vector Machines is the best classification method for fraud detection in auto insurance claims since it has the highest accuracy, recall, and F1 score values, along with a good precision value.

Insurance fraud is an essential and costly challenge for all insurance-related industries. In Indonesia, insurance claim fraud is still widespread. The number of fraudulent insurance claims has made insurance companies focus on minimizing fraud. Nevertheless, insurance companies find it challenging to manually verify each claim due to the high number of claims. Therefore, this study will use machine learning techniques to identify fraud in an auto insurance claim. In this study, fraud detection is carried out using the Bootstrap Aggregating classifier method with three base learners, namely Logistic Regression, Support Vector Machines, and Naïve Bayes—implementation of Bootstrap Aggregating on the base learner to improve model performance without overfitting. Based on the analysis conducted, it is concluded that Bootstrap Aggregating on Support Vector Machines is the best classification method for fraud detection in auto insurance claims since it has the highest accuracy, recall, and F1 score values, along with a good precision value.

Kata Kunci : Fraud Detection, Insurance, Ensemble Learning, Bootstrap Aggregating, Logistic Regression, Support Vector Machines, Naïve Bayes