Subset ARMA Model Identification for Monthly Electricity Consumption Data

Abstract

Subset models can always be highly influential in series analysis, particularly when the data demonstrate a sort of form in periodic behavior with miscellaneous natural period's ranges, specifically; days, weeks, months and years. Subset models can also be effective as they let the number of parameters lower allowing only the really needed ones to be present in the model. Though subset autoregressive moving-average (ARMA) models always receive much attention, their identification is computationally cumbersome. This paper aims at the identification of Subset ARMA model through utilizing two methods of identification; innovation regression method and genetic algorithm method. The innovation regression method is a traditional one whilst the genetic algorithm methodologies represent a relatively modern approach for identifying Subset ARMA models in recent decades. After encoding every ARMA model as a binary string in the latter method, the iterative algorithm tries tracing the natural evolution of the population in those strings through letting strings to reproduce, producing newer models for competing for survival within upcoming populations. The aim of this research is to show the procedures for identifying the most appropriate order of subset ARMA models for the monthly electricity consumption data in Damietta governorate.