In a globally competitive environment, it is necessary for enterprises to grasp the price trend of material for drawing up an effective overall manufacturing plan. However, material prices are often dramatically unstable making it difficult to obtain accurate, short-term predictions using long historical observations. Facing the challenge, forecasting with new limited data is considered more effective, efficient, and of considerable interest. Grey theory is one useful approach that can effectively handle uncertain problems under small data sets, and the AGM(1, 1) is a forecasting model based on this theory that can obtain better forecasting outcomes. However, this technique is initially suggested for use with very short-term forecasting. To improve this drawback for a longer forecasting, this study aims to extend the applicability of the AGM(1, 1) by combining the AGM(1, 1) with the rolling framework technique into a useful model referred to, here, as AGMRF. The forecasting performance of AGMRF is confirmed in this study using aluminum price data collected from the London Metal Exchange (LME). The results are compared with the other four methods, GM(1, 1), AGM(1, 1), LR, and BPN, and show that the proposed procedure can effectively deal with the problem when the sample size is limited.
|Number of pages||16|
|Journal||Economic Computation and Economic Cybernetics Studies and Research|
|Publication status||Published - 2013|
All Science Journal Classification (ASJC) codes
- Economics and Econometrics
- Computer Science Applications
- Applied Mathematics