The latest trend tracing model for projection of mineral demand |
| |
Authors: | Guocheng Pan DeVerle P Harris |
| |
Institution: | (1) NERCO Inc., 500 NE Multnomah, 97232 Portland, Oregon, USA;(2) Department of Mining and Geological Engineering, University of Arizona, 85719 Tucson, Arizona, USA |
| |
Abstract: | A new method for short- and long-term forecasting of mineral commodities based upon historical data is developed. The method, referred to as the latest trend tracing (LTT) model, is constructed as a weighting and adaptive approach based on a general linear model. The LTT model considers the functions of data location and statistical behavior. The newest data receive the largest weights, whereas the older data are given smaller weights. The LTT model is performed by an iterative algorithm. The data set is successively partitioned into training and testing subsets. Each LTT model is estimated and tested for each partition. The updated estimates are then synthesized to produce the final estimates based upon the data locations and estimation variances. The LTT model is demonstrated on two real case studies, one on the projection of U.S. aluminum consumption and the other on the forecasting of U.S. copper consumption. |
| |
Keywords: | Linear model Mineral consumption Forecasting Latest trend tracing |
本文献已被 SpringerLink 等数据库收录! |
|