| Effective supply chain management is one of the key determinants of success of today's businesses. A supply chain consists of various participating businesses that ultimately provide value to the end consumer. However, communication patterns that emerge in a supply chain tend to distort the original consumer's demand and create high levels of noise (randomness). This distortion and noise negatively impact forecast quality of the participants. In this thesis we comparatively examine the quality of new artificial intelligence (AI) based forecasting techniques to identify if they can provide increased forecast accuracy in such conditions. Experiments are performed using data from a chocolate manufacturer, a toner cartridge manufacturer, as well as data from the Statistics Canada manufacturing survey.; A representative set of traditional and AI-based forecasting techniques are applied to the demand data and the accuracy of the methods are compared. As a group, the average performance in terms of rank of the AI techniques does not outperform the traditional approaches. However, using a support vector machine (SVM) that is trained on multiple demand series as one general model, which we call a Super Wide model, has produced the most accurate forecast. Providing a large number of examples to this specific AI technique was the key to achieving high quality forecasts. This leads us to conclude that the best AI technique does outperform its best traditional counterpart, which is exponential smoothing. |
