Comparing the Performance of Seasonal ARIMAX Model and Nonparametric Regression Model in Predicting Claim Reserve of Education Insurance

Abstract

One of the biggest problems in the continuity of one's education is the education fee which is often unaffordable. Therefore, the existence of education insurance is a solution to this problem. Along with increasing public interest in education insurance, insurance companies need to adjust the claims reserves with the number of claims paid to maintain the company's capital. Claim reserves are funds that must be provided by insurance companies to fulfil obligations to policy holders in the future. Losses and inaccuracies in the payment of insurance claims will result in the policy holder and the insurance company itself. Therefore, it is necessary to do a prediction of insurance company's monthly reserve claims. In education insurance, the claim reserve data has seasonal characteristics and the number of educational insurance claims tends to increase at the turn of the school year. These fluctuating patterns are supposed to fit the application of the SARIMA model and the nonparametric regression model with the Fourier series estimator in forecasting. Fourier series is a function that has flexibility in approaching fluctuating, seasonal, and recurring data patterns. The results showed that the prediction accuracy of the SARIMAX model was higher than the nonparametric regression model with MAPE of 15% and 4% respectively.