Anti-Fraudulent System for Insurance Using Random Forest Classifier
K. Jaspin, E Anitha, Y. Jeya Sheela, A. Vinora, E. Ajitha, D Vasithra · 2024
Insurance fraud is a serious crime, placing a heavy and personal burden on individuals, businesses, government agencies, and the general public due to the density between dynamic factors such as health care providers, patients, and services. Increased incidents contribute to increased total costs incurred by insurers, premiums paid by policyholders, and states budgets This article covers a variety of impermissible practices and illegal practices around. The existing system is machine learning model vote classification (VC) with an accuracy of $\mathbf{8 6 \%}$ for anti-fraudulent detection. The proposed system is random forest classification algorithm using machine learning algorithms to detect fraudulent activities. The machine learning algorithms designed to predict whether an individual has made a fraudulent or legitimate insurance claim. First, duplicates and null values are removed from the data set. This method divides the collected data into several decision trees according to attributes such as gender, insurance coverage, frequency of employer’s insurance premiums and then displays a graph representing the divided data for data analysis. The Random Forest Classifier is specifically used for detecting fraudulent insurance policies. After classification, the results are summarized in a classification report, and a confusion matrix is generated to present the detection results for both fraudulent and non-fraudulent claims. The new system achieves an accuracy score of $\mathbf{9 0. 0 0 8 \%}$, demonstrating a robust ability to detect most fraudulent claims accurately.