Big data and machine learning-based decision support system to reshape the vaticination of insurance claims

Abstract

Based on actuarial science theory, decision-making theory, and anonymous big data, this study employs machine learning to advance insurance claim forecasting, aiming to enhance pricing accuracy, mitigate adverse selection risks, and optimize operational efficiency for improved customer satisfaction and global competitiveness. The study utilized the Boruta algorithm with LightGBM for feature selection, analyzing a 57-dimensional dataset and identifying an optimal subset of 24 features. The improved LightGBM model achieved superior results (AUC ∼ 0.9272 and accuracy ∼ 92.94 %), surpassing other models evaluated. Beyond operational improvements, the proposed model holds the potential to contribute to various United Nations SDGs, such as promoting financial inclusion (SDG 1; SDG 10), reducing fraud, improving public safety (SDG 3; SDG 11; SDG 13), and encouraging sustainable practices (SDG 9; SDG 11). By utilizing data-driven insights to make more informed and accurate decisions, insurance companies can provide better services to their policyholders and contribute to a more equitable and sustainable society.