Using neural networks to monitor supply chain behaviour

Abstract

Intelligent agents are expected to play an increasingly important role in Supply Chain Management (SCM) by automating event-tracking, trend-prediction and decision-making functions. In this paper, we proposed a new trend-prediction methodology that recognises behavioural patterns and predicts future performance based on those patterns. We used fuzzy Adaptive Resonance Theory (ART) Neural Networks (NNs) to build the patterns and BackPropagation NNs (BPNNs) to make the predictions. We based this methodology on System Dynamics (SD) models, which were used to train the NNs. We believe that our approach could be incorporated easily into a number of software agents. These agents could improve dramatically the capabilities of current dashboard-monitoring systems.