Information Technology Reference
In-Depth Information
the flexibility of the supply chain (Gossain, Malhotra, & El Sawy, 2005). Gossain et al. (2005) argue that
developing robust and reconfigurable links would promote the agility of the chain in terms of offering
and partnering flexibilities. In their study they found that while the quality of the information sharing
in a supply chain could promote flexibility, the breadth of information shared has a detrimental effect
on it. The modularity and loose couplings between the partners have been identified as positive factors
in this regard. Overall, we see many realities that effectively form information exchange collaboration
barriers, which limit the possibilities of information exchange within the supply chain.
Effective demand forecasting is therefore still a serious hurdle for many businesses. The objective of
this work is to study the feasibility of forecasting the distorted demand signals in the extended supply
chain using the advanced machine learning techniques as opposed to the more traditional techniques at
the upstream (manufacturer's) end of the supply chain. We are particularly interested in the upstream end,
since that is where the distortion is at its worst, and the demand swings tend to be most erratic. In light
of the above considerations, the problem of forecasting distorted demand is of significant importance
to businesses, especially to those operating towards the upstream end of the extended supply chain.
In this work we investigate the potential value of applying advanced machine learning techniques,
including artificial neural networks (ANN), recurrent neural networks (RNN), and support vector ma-
chines (SVM) to demand forecasting in supply chains. These learning techniques permit machines to
identify patterns in data. The performances of these machine-learning (ML) methods are contrasted
with baseline traditional approaches such as exponential smoothing, moving average, linear regression
and the Theta model. To this end we have collected real industry data from three different sources. The
first two data sets are from the enterprise systems of a chocolate manufacturer and a toner cartridge
manufacturer. Both of these companies, by the nature of their position in the supply chain, are subject to
considerable demand distortion. The third source of data comes from the Statistics Canada manufactur-
ing survey. Inclusion of this survey in the study has the aim of increasing the validity and facilitating
the possibility of replication of results by others.
In the sections that follow we provide the background, describe the data sources and the experimental
setup and present the results of our experiments. The work concludes with the discussion of findings
and directions for future research.
Background
This section discusses the problems with demand distortion in supply chains, and reviews traditional
and machine learning-based (ML-based) forecasting techniques.
demand distortion in Supply Chains
Companies need to consider the inter-relationships among demand forecast, perfect order, and inventory.
Hofman (2007) indicates that there is a strong correlation between demand forecast accuracy and the
perfect order and that companies that are better at demand forecasting have significantly lower inven-
tories, stronger perfect order fulfillment, and shorter cash-to-cash cycle times. She argues that demand
planning and forecasting methods and their supporting technologies enable or constrain a company's
business performance metrics. Accordingly, as indicated by (Raghunathan, 1999), one of the major
purposes of supply chain collaboration is improving the accuracy of forecasts.
Search WWH ::
Custom Search