Performance Analysis of Classifiers in Predicting Car Insurance Claim

Insurance claims pose a significant and costly challenge for insurance companies. Managing insurance claims presents a fundamental issue for insurance firms. Insurers continually face the challenge of mitigating the escalation of claim losses due to fraudulent activities and the increasing volume of claim data within their operations. Consequently, accurately assessing the status of insured claims during the review process becomes increasingly complex. In order to make the car insurance claim prediction more accurate and feasible a comparison study on the performance of various machine learning and deep learning algorithms is presented in this paper. In this research study, the primary research contribution is that various machine learning and deep learning algorithms like logistic regression, decision trees, random forest, catboost, multilayer perceptron etc. have been analyzed and performance has been evaluated based on the accuracy, precision, recall, F1-score, ROC AUC and confusion matrices. This study shows that some algorithms outperform the others and the Random Forest (RF) algorithm has resulted with the highest accuracy of $\mathbf{9 8. 0 \%}$.