Vehicle Insurance Fraud Detection Using Supervised Neural Network Model

Abstract

Vehicle insurance fraud is a serious problem that causes huge financial losses to insurance companies. In recent years, developing fraud detection models using machine-learning techniques has been of great interest. This study proposes a novel approach for vehicle insurance fraud detection using Supervised Neural network model on the Kaggle Vehicle Insurance Fraud Detection dataset. The initial step in this study involves performing data pre-processing and feature extraction on the dataset, which consists of information such as ages of those insured, the gender of the policyholders, and the make and model of the vehicles. As traditional rule-based techniques are not adequate for detecting fraudulent claims, there are ongoing investigations on machine learning-based methods. Our research proposes a supervised learning method using a neural network model for vehicle insurance fraud detection. The proposed model is trained on a large, publicly available dataset and assessed using standard evaluation metrics such as F1 score, recall and precision. Moreover, we compare the model’s performance with other well-known machine learning algorithms such as decision trees and logistic regression. Our exploratory results show that the proposed approach gains increased accuracy in detecting fraud cases. The proposed learning algorithm outperforms other traditional supervised learning algorithms in terms of classification performance, demonstrating the effectiveness of the proposed approach. Overall, our suggested approach shows a promising solution for vehicle insurance fraud detection, with the potential to reduce financial losses significantly for insurance companies.