Robust and resourceful automobile insurance fraud detection with multi-stacked LSTM network and adaptive synthetic oversampling

Le résumé fourni par la source

Insurance companies worldwide are concerned about financial losses due to false claims. Automobile insurance fraud (AIF) has become more sophisticated, causing the yearly loss of trillions of dollars. AIF is tough to establish, and acquiring a thorough knowledge of the problem is complex. Also, AIF investigators have relied on manual claims inspection, proving costly, inefficient, and time-consuming. This paper proposed a robust and resourceful approach to AIF detection with multi-stacked LSTM (MSLSTM) reinforced with the adaptive synthetic (ADASYN) sampling algorithm for imbalanced learning. We experiment with the proposed model with a publicly available AIF dataset from Kaggle. Using accuracy, recall, precision, F1-score, and AUC, we compared the performance of our proposed MSLSTM model with well-known machine learning algorithms and previous AIF detection works. Our results showed a fair performance (accuracy = 95%, precision = 94%, AUC = 97% and F1-score = 92%) of the MSLSTM model than other algorithms and works.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.