Comparison of ARIMA and ARIMA/GARCH Models in EVN Traffic Prediction

  • Tran Quang Thanh
  • Trinh Quang Khai

Abstract

This  paper  focuses  on  building  statistical models  to  capture  and  forecast  the  traffic  of  mobile communication  network  in  Vietnam.  Following  BoxJenkins  method,  a  multiplicative  seasonal  ARIMA model is constructed  to  represent  the  mean  component using the past values of traffic, a GARCH model is then incorporated  to  represent  its  volatility.  The  traffic  is collected  from  EVN  Telecom  mobile  communication network.  The  numerical  result  comparisons  show  that the  multiplicative  seasonal  ARIMA/GARCH  model built  in  this paper gives a better estimate  when dealing with volatility clustering in the data series. However, in short-term  prediction  where  the  volatility  has  an insignificant influence, the achieved ARIMA model also can  be considered  as  a  good  model  to  capture  well  the characteristics  of  EVN  traffic  series  and  gives reasonable forecasting results.
Published
2014-10-28
Section
Regular Articles