Motor Insurance Claim Processing and Detection of Fraudulent Claims Using Machine Learning

  • Mariya Mathew, Nimitha M Kunjumon, Ria Maria Lalji, Kency Susan Skariah, Dr Jeyakrishnan V


A claim that is poorly handled will bring criticism through social media and hence detection of fraudulent claims are highly alarmed in the society. The paper focuses on estimating the insurance amount provided to the customer based on the severity of vehicle damage during an accident and detecting fraudulent claims. Random Forest is used to build a regression model that might be applied in forecasting the insurance claim. The algorithm involves identification of associations between claims, high dimensionality application to cover all levels, identification of missed observations, etc. By that way, the portfolio is made for the particular customer. An automobile insurance company's actual data was chosen to create the random forest model centered on the automotive insurance fraud removal theory. The data’s were analyzed, and the dependency of each input variable to the corresponding output variable was obtained. The error of each model was examined. Finally, the model was tested by empirical study. The empirical findings indicate that: the automotive insurance fraud mining paradigm that incorporates Random Forest is ideal for wide data sets and unstable data relative to the traditional model. It can be well used to predict the claim amount to be given to the insured person and to detect the fraudulent claim of insurance.