AI is becoming seen as the solution to every problem including demand planning. The Centre for Marketing Analytics and Forecasting hosted two free events on this topic to discuss the latest research advances using Artificial Intelligence with the tutorial providing the opportunity to engage in depth in developing their understanding of AI’s potential. Both well-attended events brought together practitioners from industry and academics discussing this hot topic.
The half-day workshop consisted of three talks. First Sven Crone, from Lancaster University Management School, provided a talk on how artificial intelligence is changing demand planning and inventory control today. His talk gave examples of how industry leaders have successfully implemented artificial intelligence, machine learning and data science for forecasting in supply chain management and logistics. He also stressed the point that AI can go beyond simple forecasting task and provide useful insights on tasks such as the identification of outliers from complex time series or to learn cost-efficient inventory levels.
In the second talk, Ralph Grothmann, Principal Expert for Predictive Analytics at Siemens, presented an application of how to forecast customer demand with deep neural networks. The case looked at feedforward neural network forecasts for more than 1000 products. The solution is now integrated into the ERP system of the Siemens factory demonstrating significant improvements over their established machine learning methods.
The last talk was by David Salinas, Senior Machine Learning Scientist at Amazon AI Labs. He discussed deep learning for forecasting at Amazon: problems and methods. The various case studies and examples of forecasting problems highlighted the strength machine learning methods have as well as their flexibility: they can deal with many hundreds of thousands of products, some with lots of data and regular sales, others with intermittence and many zero sales periods, and even new products. He also explained what skills and software are needed to tackle these problems at scale.
Our one-day tutorial aimed towards industry practitioners in the area of demand planning and predictive analytics to provide them with the first hands-on experience in how to forecast with Artificial Intelligence using neural networks. The course elaborated on how to apply simple neural networks to various forecasting tasks and understand the benefit of applying deep learning methods. In the morning session, Sven Crone provided insights on how to use shallow feedforward Neural Networks for Forecasting. The afternoon session then included a tutorial on Deep and Deep Recurrent Neural Networks taught by Ralph Grothmann.