<![CDATA[Generalized Space Time Autoregressive (GSTAR) models are generalization of the Space Time Autoregressive (STAR) models which has the data characteristics of time series and location linkages (space-time). GSTAR is more flexible when faced with the locations that have heterogeneous characteristics. The purposes of this research are to get the best GSTAR model and the forecasting results of Consumer Price Index (CPI) data in Purwokerto, Solo, Semarang and Tegal. The best model obtained is GSTAR (11) I(1) using cross correlation normalization weight because it generated white noise and multivariate normal residuals with average value of MAPE 3,93% and RMSE 10,02. The best GSTAR model explained that CPI of Purwokerto is only affected by times before, it does not affect to other cities but can be affecting to other cities. Otherwise, CPI of Surakarta, Semarang and Tegal are affecting each others. Keywords: GSTAR, Space Time, Consumer Price Index, MAPE, RMSE]]>
Copyrights © 2015
