EXPLAINABLE AND OPTIMIZED MACHINE LEARNING FOR VEHICLE INSURANCE FRAUD DETECTION USING ABC – XGBOOST- SHAP INTEGRATION

Abstract

Effective detection of insurance fraud is essential for the financial sector, requiring advanced methodologies that ensure accuracy and transparency. This study introduces a novel framework that integrates an XGBoost classifier with the Artificial Bee Colony (ABC) algorithm and the SHAP (SHapley Additive exPlanations) approach for enhanced insurance fraud detection.The ABC-XGBoost with SHAP framework optimizes feature selection and hyperparameter tuning, leveraging the strengths of both the ABC algorithm and XGBoost. Experimental evaluations on real-world insurance datasets indicate that this approach outperforms traditional fraud detection methods, achieving higher accuracy and reduced false positive rates.This research advances the field of insurance fraud detection by demonstrating the effectiveness of merging machine learning and optimization algorithms, while providing transparency through the insights offered by SHAP values.