Actuarial Risk Assessment and Pricing Optimization Algorithm Based on Graph Neural Network

In this study, we proposed a new method combining graph sampling aggregation network (GraphSAGE), Contrastive Learning (CL) and Bayesian Optimization. First, GraphSAGE was used to model the graph structure of risk data, capture the complex relationship between nodes, and provide the basic feature representation for risk assessment. Secondly, contrast learning is used to enhance the model's ability to learn risk features, maximize the similarity of positive sample pairs and minimize the similarity of negative sample pairs, and further optimize the feature representation. Finally, Bayesian optimization is introduced to dynamically adjust the model parameters, optimize the pricing strategy, and improve the adaptability of the model to different risk scenarios. Experiments show that CBEGN performs well on multiple actuarial risk datasets. Compared with traditional methods, the proposed model can more accurately identify complex relationships between nodes, optimize pricing strategies, reduce errors, and maintain high efficiency in the face of novel data analysis tasks or previously unseen data. This fully proves the effectiveness and feasibility of CBEGN in actuarial risk assessment and pricing optimization.