Forecasting Model Selection for Time Series Using Global Characteristics and Neural Network

摘要：

Forecasting model selection is an important component of model management of decision support system (DSS). A crucial problem confronting users of DSS is that the present methods of forecasting model selection require the users too much. The users are required to have a profound understanding of the models in the model base and the problems to be solved. This leads to poor feasibility of forecasting model selection. In this paper we test the feasibility of employing the neural network structure for model selection. To accomplish this objective, global measures describing the time series are obtained by applying statistical operations that best capture the underlying characteristic: trend, seasonally, periodicity, serial correlation, skewness. kurtosis. chaos, nonlinearity. and self-similarity. Since using extracted global measures, it reduces the dimensionality of the time series and is much less sensitive to missing or noisy data. A back propagation neural network is then constructed with thirteen input nodes representing time series characteristics. The output nodes of the neural network represent three categories time series forecasting methods. The results indicate that using global characteristic and the neural network approach can assist the practitioner in the selection of the appropriate forecasting model.

关键词： Forecasting model selection Neural network Decision support system Global characteristic

作者: Dabin Zhang Chong Chen Huiyuan Tu

作者单位: Information Management Department, Huazhong Normal University, Wuhan 430079, P.R.China Institute of Institute of Systems Science, Academy of Mathematics and Systems Science, Chinese Academy of Science

会议类型: 国际会议

会议名称: The First World Congress on Global Optimization in Engineering & Science(第一届工程与科学全局优化国际会议 WCGO2009)

会议地点: 长沙

会议语种:英文

页码: 874-879

在线出版日期: 2009-06-01（万方平台首次上网日期，不代表论文的发表时间）