Comparison of forecasting models for first year premium of ordinary insurances by holt-winters exponential smoothing, box-jenkins and artificial neural networks methods
Abstract:
The objective of this research was to study forecasting models for the life insurance first year premium of ordinary product. The premium data was gathered from the Office of Insurance Commission (OIC) during January, 2003 to December, 2018 of 192 VALUes were used and divided into 2 Sets. The first set of data from January, 2003 to December, 2017 was applied for constructing forecasting models by Holt- Winters Exponential Smoothing Method (HWS), Box- Jenkins method and Artificial Neural Networks and employed Root Mean Square Error (RMSE) as the criteria for model selection. The second set of data from January, 2018 to December, 2018 Was used to measure the performance of forecasting model by using Mean Absolute Percentage Error (MAPE) to show the percentage error between real data and forecast value. The result showed that the forecasting model by Holt-Winters Exponential Smoothing Method was the appropriate method and had the highest forecasting accuracy which gave the minimum RMSE and MAPE was 15.55% when employing this method to forecast the life insurance first year premium of ordinary product.