SARIMA-ELM hybrid model versus SARIMA-MLP hybrid model

Abstract

Advanced quantitative models have been widely developed in academic literature, but practicians and statisticians still have little interest in the design and development of sophisticated hybrid models to produce more accurate predictions. In an empirical study, we compare performance forecast accuracy of the SARIMA-ELM hybrid model and the SARIMA-MLP hybrid model for forecasting tourist arrivals to Bali from 10 different countries such as China, Australia, Japan, India, USA, UK, Canada, South Africa, Malaysia, and Taiwan. Based on the RMSE and MAPE criteria, we found that our novelty model (the SARIMA-ELM hybrid model) performs better than the SARIMA-MLP hybrid model in the aspect of forecast accuracy.