Siemens, Munich Germany
Forecasting Customer Demand with Deep Neural Networks
Accurate forecasts of the customer demand are key for a successful supply chain management. In the production process, the material and capacity can be planned on the basis of the expected demand. Delivery capability and reliability can be ensured. We apply deep feedforward neural networks to explore demand patterns in the sales time series of more than 1.000 products. The models use autoregressive as well as seasonal components. In addition macro-economic factors are used to explain the demand fluctuations. The approach incorporates an automated model building, training and evaluation scheme. The forecasts are integrated into enterprise resource planning system of a Siemens factory. We benchmark the forecast accuracy of our approach with state-of-the-art machine learning methods.
Studies of economics in Bremen with a focus on finance & investment analysis, portfolio management and econometrics, graduation 1999 in economics. 2003, PhD in economics, University of Bremen. Title of the PhD-thesis: ‘Multi-Agent Market Modeling based on Neural Networks.” Since 2003 at Corporate Technology, Siemens AG, Department of Business Analytics and Monitoring, Munich. Since 2018 with Siemens Germany, Digital Enterprise & Digital Services. Current position: Principal Consultant for Predictive Analytics. Current research interests: Applications of time-delay recurrent and feedforward neural networks. Member of the scientific advisory board of the German Society of Operations Research (GOR).