Information Technology Reference
In-Depth Information
However, full collaboration is not always possible and therefore it is important to investigate the
feasibility of forecasting demand in the absence of extensive information from other partners. The
source of the demand distortion in the extended supply chain is due to demand signal processing by the
members in the supply chain (Forrester, 1961). According to Lee, Padmanabhan et al. (1997b), demand
signal processing means that each party in the supply chain does some processing on the original de-
mand signal, thus transforming it before passing it along to the next member. As the end-customer's
demand signal moves up the supply chain, it becomes increasingly distorted. This occurs even if the
demand signal processing function is identical in all parties of the extended supply chain. For example,
even if all supply chain members use a 6-month trend to forecast demand, distortion will still occur.
The phenomenon could be explained in terms of the chaos theory, where a small variation in the input
could result in large, seemingly random behavior in the output of the chaotic system, which, in the
context of the supply chain leads to the “bullwhip” effect. Figure 1 depicts a section of the supply chain
with a collaboration barrier, i.e., the link at which no additional information sharing occurs between
the partners, which results in distorted demand forecasting. Our objective is to forecast future demand
based only on manufacturer's past and current orders. An increase in forecasting accuracy will result
in lower costs because of reduced inventory as well as increased customer satisfaction and retention
that will result from an increase in on-time deliveries (Stitt, 2004).
It has been shown that the use of simple techniques such as moving average or naïve forecasting will
induce the bullwhip effect (Dejonckheere, Disney, Lambrecht, & Towill, 2003), while autoregressive
linear forecasting could diminish it (Chandra & Grabis, 2005). Furthermore, a simulation based study
has shown that genetic algorithm-based artificial agents can achieve lower overall costs in managing
supply chain than human players can (Kimbrough, Wu, & Zhong, 2002). In this work we will evaluate
the performance of advanced machine learning techniques to investigate their applicability and compare
their performance with the more traditional forecasting methods.
Traditional f orecasting Techniques
Extensive research on forecasting has provided a large number of forecasting techniques and algorithms in
mathematics, statistics, operations management and supply chain academic outlets. Forecasting competi-
tions have consistently found that the simpler forecasting methods had better overall accuracy than more
complex ones (Makridakis, Andersen, Carbone, & Fildes, 1982; Makridakis et al., 1993; Makridakis &
Figure 1. Distorted demand signal in an extended supply chain
Distorted Demand Forcasting
Manufacturer's Distorted
Demand
Collaboration
Barrier
Forecasted Future
Demand
Past Demand
Data Mining
Order
Manufacturing
Distributor
Decreased inventory
Increased on-time deliveries
Search WWH ::
Custom Search