A review on the combination of autoregressive forecasting models in multiple time series

In the recent years, hybrid strategies have received much attentionbecause they reduce the potential errors due to forecasting and achievegood performance. As the use of these strategies and neural networks topredict large and voluminous data has been extended, there ischallenges of their simultaneous application. M۴ forecastingcompetition, which has been implemented in recent studies, is thecontinuation of the three previous matches. This began more than ۴۵years ago and aimed to learn how to improve foresight and use suchlearning to advance theory and practice and indicate the superiority offorecasting methods over each other. This paper first reviews severalresearches related to forecasting strategies from ۲۰۱۸ and reviewsseveral articles associated with the M۴ forecasting competition.

Keywords:

Forecast combinations , M۴ prediction competition , Neural networks , Time series