Modeling Vehicle Insurance Adoption by Automobile Owners: A Hybrid Random Forest Classifier Approach
Moin Uddin, Mohd Faizan Ansari, Mohd Adil, Ripon Kumar Chakrabortty, Michael J. Ryan · Processes · 2023
This study presents a novel hybrid framework combining feature selection, oversampling, and machine learning (ML) to improve the prediction performance of vehicle insurance. The framework addresses the class imbalance problem in binary classification tasks by employing principal component analysis for feature selection, the synthetic minority oversampling technique for oversampling, and the random forest ML classifier for prediction. The results demonstrate that the proposed hybrid framework outperforms the conventional approach and achieves better accuracy. The purpose of this study is to provide insurance managers and practitioners with novel insights into how to improve prediction accuracy and decrease financial risks for the insurance industry.