Abstract:
This research aims to compare efficiency of forecasting model for time series when
the variance of Heteroscedastic is inconstant which have three ways are; First,
Autoregressive Conditional Heteroscedasticity (ARCH) Second, Smooth Transition
Autoregressive (STAR) and Third, Transition Autoregressive (TAR). Time series
simulation method is used when the variance is inconstant by using Autoregressive
Conditional Heteroscedasticity (ARCH). Coefficient parameter of model and sample
size are specified. The n of this research are 60, 200, 500, 800, and 1,000 and it was
found that the different of data sizes would have efficiency in forecasting indifferently.
However, the efficiency in forecasting of n=60 is different from the other groups.
Furthermore, the parameter α0 at 0.1, 0.5, and 0.9 and α1 at 0.5 and 0.9 was found that
parameter in different level would be conformed except parameter α0 at 0.9 and
parameter α1 at 0.9.