Projection of demand with recurrent neural networks
In production and operations management, forecasting demand is an important method, as it helps to develop better approximations of future operations in the presence of uncertainty. Forecasting extracts mathematical relationships from past data that can be used to inform future decision making. In supply chain management, efficient coordination of resource acquisition, production and warehousing is highly dependent on accurately predicting future demand for particular products and overall market dynamics. Accurate forecasting of demand reduces investment risks in uncertain environments.
The challenges of forecasting demand lie in the complexity of the dynamics of demand. We investigated the application of reservoir computing (RC) to forecasting product demand. RC uses a randomly initialized recurring neural network that implements finite memory and generalization. Under these conditions, it should be sufficient to reduce the complexity of the training to a single linear output layer and achieve accurate forecasting results. Therefore, the output layer is capable of deriving a simple linear relationship between the input data and its projection into a larger dimension feature space.
Project manager:
Jens Burger, PhD
Email: jensburger@upb.edu
Project duration:
July 2018 - June 2019
