Accuracy of Periodogram Analysis for Identification of Multiplicative SARIMA Models
Gumgum Darmawan, Nuning Kurniasih, Ansari Saleh Ahmar · Journal of Physics Conference Series · 2018
In everyday life, seasonal events are common, seasonal behavior is common in various periods, such as daily, weekly or monthly. business, economics and financial cases for example, we often encounter a seasonal phenomenon, namely data that repeated over the same period. However, this seasonal test is only accurate For stationary Seasonal time series data. So, in this research we apply exact identification for multiplicative time series data from generated data. Performance of this test is determined by the percentage of fit identification. We apply this Periodogram Analysis to real data, AirPassengers and UKgas data. All simulation and real data analysis were done by open source Software R (OSSR). It is observed that in all model for Period 12, accuracy of detection seasonal models under 50%. However, for P=6, The algorithm was accurate enough to detect multiplicative-seasonal models, 70% of seasonal model can be detected by the algorithm.