Analysis And Modeling Method On Short Time Series

Along with the flourishing development of the high-tech and the Internet industry of China, large amounts of data are generated by production and management of the government and enterprise. Using these data to scientifically improve accuracy of administrator's decision-making is a significant means of enhancing government's governing capacity and enterprise's market competitiveness. Time series analysis, as an important branch of mathematical statistics, is a technique to find the change rule of time series and forecast its trend of future.Short time series coupled with few historical observations largely come from the process of production and management. Therefore, it is valuable to analyze short time series. However, the traditional methods of time series analysis are not fit to analyze short time series and so far there are not relatively mature modeling system for short series. Furthermore, the collection for the long series needs more human resource and technology support compared with the short series. This will increase cost for manufacturing management. Therefore, it is significant to propose an effective model for forecasting short time series. This paper propose a model to forecast short time series. The method proposed by this paper outperforms ARIMA when stable seasonal patterns and only a small amount of data is available. The modeling procedure applies Bayesian method.Real example from the data of aviation passenger will be analyzed by the proposed procedure. The Analytic results will compare with ARIMA's. This paper also proposes an adjusting method for forecasting the set of short time series. This procedure will be applied by the example of airport substation.