AUTO-CORRELATION OF DEMAND IN A MULTI-ECHELON SYSTEM.
Abstract
The paper considers the question of whether or not inventory levels at lower echelons should be used in the forecasting design for a multi-echelon supply system. The conjecture had been made that maintaining inventories (maximum assets) at the lower echelons proportional to their demand forecast would produce sufficient time dependence in the upper echelon demand pattern to result in demand autocorrelation at that level large enough to make possible a significant improvement in forecasting. Under the assumptions that (1) demand at the bottom is an independent wide-sense stationary stochastic process, (2) single exponential smoothing is used for demand forecasting, and (3) infinite demand history is available, upper echelon demand variability and autocorrelation coefficient function, R(n), is computed for several cases that vary in the degree of supply system complexity. In all cases R(n) is always negative and the absolute value is (a) bounded in the interval (0, .5), (b) monotonically decreasing with increasing time lag n, (c) monotonically increasing with increasing value of the smoothing constant. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 15, 1967
- Accession Number
- AD0666786
Entities
People
- Steven Gajdalo