An Analysis of Forecasting Techniques for Wholesale Demand: The Applicability of Multi-Model Forecasting.
Abstract
This study focuses on determining if a multi-model forecasting strategy produces a more accurate demand forecast than the present eight-quarter simple moving average used in the Air Force Logistics Command D062 inventory control system. The analysis uses ten years of actual expendable (non-recoverable) data. This research analyzes the following forecasting techniques: naive, simple moving average (4, 8, and 12 periods), double moving average (4, 8, and 12 periods), single expontential smoothing (alpha of 0.2 and 0.8), single expontential smoothing with trend, focus forecasting, simple regression, S-curve analysis, expontential growth and eclectric methods. The analysis compares the fifteen techniques in terms of the mean absolute deviation (MAD) and percentage change, tracking signal, and variance. Also, the statistical test, Oneway Analysis of Variance (ANOVA) compares the forecasting technique results, The results show simple expontential smoothing with an alpha of 0.2 as the forecasting model with the lowest MAD and variance. The techniques of simple moving average (4 and 8 periods) and simple expontential smoothing (alpha=0.8) exhibit very similar results. The ANOVA test shows no significant difference between the eclectric method and the top twelve techniques. However, from the other test results, the eclectic method and focus forecasting performed somewhat inferior for the data tested. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1982
- Accession Number
- ADA122884
Entities
People
- Daniel Lee Gartner
- Grace Ann Bittel
Organizations
- Air Force Institute of Technology