Published April 2019.
Schlötter provided SDSC-BW’s data analytics experts with sales data from the past 13 years: a total of about one million data elements. The data included sales information for each product on each day, such as sales volume, warehouse address, customer number, order time, delivery time, etc.
To reduce the losses caused by inaccurate forecasts, the SDSC-BW team tried to use machine learning to optimize the sales volume prediction. It considered the problem for this purpose as a time series prediction task. Since under-prediction can lead to a threefold higher loss compared to over-prediction, the experts defined an asymmetric evaluation metric.
Data Innovation Community
Schlötter, Smart Data Solution Center Baden-Württemberg
April 2019 – October 2019