A Survey on Property Insurance Claims Using Machine Learning Models in Finance Sector

Abstract

The complexity of property insurance claims and the rising number of risk factors involved in financial risk prediction are factors that have made proper early claim prediction an essential requirement among insurers. Conventional means of assessment cannot process heterogeneous data, and increases of fraud attempts and the changing conditions of the market, which demonstrates the need to use more efficient and data-focused approaches. This paper considers the property insurance terrain within the finance industry and investigates how supervised and unsupervised machine learning methodologies can be used to enhance the accuracy of claim prediction and the operational decision-making process. The survey summarizes the important considerations in claims, key important ML approaches applied in risk assessment, and issues in early prediction, such as data imbalance, scarcity of early information, and large-scale heterogeneous data. Combining the results of recent research, this paper enables systematic knowledge on the effects of ML in improving claim assessments, fraud detection, and risk modeling in property insurance. The results also bring out the existing constraints that should be overcome in order to attain credible and scalable early claim prediction.