Underwriting Decision Analysis Based on Autoregressive Moving Average Models and Random Forest Algorithms

This paper presents an analysis of underwriting decisions based on autoregressive moving average models and random forest algorithms. To begin, we quantified the influencing factors using Poisson and log-normal distributions and established a risk prediction model using multi-objective linear programming for evaluation, ultimately reaching a decision. A comparison between the model’s decision results and those of the Japan North-West Mutual Aid Bank from 2010 to 2016 showed a consistency of 78.64%.Subsequently, we constructed an autoregressive moving average model for prediction. Comparing the model’s predictions with those of a three-year insurance plan in Chile in 2010, the model achieved an accuracy of 81.32%. Finally, to investigate whether customers would influence the insurance company’s decisions, we quantified customer-related factors and utilized a random forest algorithm to establish an evaluation model, determining the impact of customers on the insurance company’s decisions.