Determine the Best Models for Time Series by using a New Suggested Technique

Abstract

The proposed method relies on a technique to choose the best model by giving values for the ranks of the model ARMA (p, q), where (p, q) are given the values 0, 1, 2. Every time (ACF) and (PACF) for the series of estimated errors {at} are tested ، and the model which satisfies the two inequalities (7) and (9) is the best. It was applied to (14) annual series, representing the amounts of rainfall, and (7) daily series representing the daily Coronavirus infections. The most important conclusions and recommendations are. Applying the proposed method reduces the possibility of committing errors because its steps are clear and straightforward, giving it an application advantage. In many cases, it is impossible to know the model's rank through the behavior of (ACF) and (PACF), while the proposed method overcomes these forms to choose the best. The research recommended adopting the proposed method because it is easy to understand, learn and apply in addition to its accuracy.