Vehicle Insurance Fraud Detection System Using Robotic Process Automation and Machine Learning

摘要

Vehicle owners purchase insurance policy for their vehicles to adjust the expenditure incurred in getting into an auto accident. Annual premiums need to be paid by owners to an auto insurance company on regular basis. The insurance company pays all or most of the costs associated with vehicle damage. The insurance sector is a highly regulated industry, and with this increasing competition, it is not at all easy to keep up with the latest technologies. Robotic Process Automation (RPA) is a promising approach that automates recurring human tasks using software bots. Many organizations use RPA to allay their employees from repetitive and tedious tasks which contributes to achieve many benefits including better business efficiency, larger productivity, data security, reduced cycle time, and improved accuracy. The client’s satisfaction, enhanced and efficient production can be achieved in the organizations with the use of computers for doing repetitive tasks which run in background along with the production processes. These background automation systems are referred as Robotic Process Automation (RPA). It automates the repeating tasks thereby reducing human intervention. RPA is known as a catalyst for the bot revolution. Implementing RPA is a challenge for any organization that is willing to adapt it and must learn to deal with RPA to reach maximum results. Usage of Robotic Process Automation (RPA) in the insurance sector facilitates the easy collection of policyholder’s details, essential information from previous years’ claims documents, therefore allowing the insurers to settle insurance claims seamlessly. This paper aims to perform Vehicle Fraud Detection by efficiently adopting Robotic Process Automation in the insurance sector to automate the task of by integrating with Machine Learning (ML) techniques that make the system more intelligent to classify an insurance claim as a fraud or legitimate. The authors found that Linear Discriminant Analysis (LDA) shown prominent results with an accuracy of 90% compared to other techniques. Finally, future directions are presented.