Simulation approach to improve performance of the seasonal time series decomposition method
Part of : WSEAS transactions on business and economics ; Vol.4, No.6, 2007, pages 87-94
Issue:
Pages:
87-94
Author:
Abstract:
A simulation method to model seasonal time series is developed. Using the ratio of seasonal data to nonlinearperiodic trend, simulation parameters of the seasonal indices are identified. A decomposed seasonal timeseries is re-seasonalized to identify the combined random as well as cyclical indices in the seasonal model.These indices are modeled as a multi channel queue system. The seasonal time series is constructed as amultiplicative decomposition simulation model. To improve performance of the time series, the simulationmodel identifies the best seasonal indices and combined cyclical and random indices. These indices areupdated whenever the simulation search improves squared error criteria of the fitted model. The simulatedoptimized seasonal, random and cyclical indices are used finally to re-seasonalize the data for forecastevaluation. The software is developed to test the performance of the approach using airline travel dataavailable in Box, Jenkins and Reinsel (1994). The proposed simulation method achieves 38.89%improvement in squared error measure against the X-11 decomposition method.
Subject:
Subject (LC):
Keywords:
simulation, seasonal, time series, indices, multiplicative, decomposition
Notes:
Περιέχει γραφήματα, πίνακες και βιβλιογραφία
