Fraud Detection in Automobile Insurance based on a Deep Hybrid Approach

Insurance fraud causes massive financial damages to vehicle insurance companies every year. Despite the many advances in insurance fraud detection, the costs created for automobile insurance companies due to such frauds are still increasing. Therefore, the main goal of this work is to present and apply different methods with unbalanced dataset in the field of predicting the occurrence of fraud in automobile insurance. In the proposed approach, we propose a new hybrid model based on deep neural network and machine learning using a data-driven approach. A proposed approach is presented for the first time to predict car insurance fraud with the goal of greater accuracy. In addition, the feature selection technique with XGboost algorithm, feature extraction with multi-layer perceptron neural network and classification with SVM algorithm are used in the field of predicting the occurrence of fraud in automobile insurance. The obtained results show the effectiveness of our proposed hybrid method in accuracy and precision of ۹۶.۹۷% and ۹۷.۰۵%. The numerical experiments demonstrate that the proposed approach achieves promising results for detecting fake accident claims.

کلیدواژه ها:

Machine learning ، Credit card fraud detection ، Artificial Neural Networks ، Random Forests ، XGBoost