Predicting Fraudulent Motor Vehicle Insurance Claims Using Data Mining Model
Jacob Muchuchuti, Stewart Muchuchuti · 2021
In recent years, insurance fraud detection has attracted a great deal of concern and attention as the insurance industry has witnessed an increase in the number of fraudulent claims. The aim of this research was to develop a data mining model that that would predict fraudulent insurance claims. The research also seeks to establish the variables with the most predictive power. From a data set with 1,000 claims, 70 percent were used for training the machine using Python Programming Language and the remainder being used to test the accuracy of the algorithms used. Twenty-nine attributes were observed to have influence in predicting the potentially fraudulent behavior of the claims and variables such as policyholder’s hobbies and the extent of damage on the motor vehicle were found to have the most predictive influence. As for the predictive influence, the Decision Tree had the highest accuracy level compared to the other algorithms.