LARGE SCALE BAYESIAN INFERENCE OF RETAIL ARTICLE WEIGHTS
Thursday 15th
from 11:30 pm to 12:15 pm
Room 19
-
Technical
Knowing the weights of individual articles in Zalando's stock is important for a number of logistics processes and operations in our warehouses. Surprisingly, this data is not provided by the suppliers in great generality or with the required accuracy. The obvious approach to manually weigh all articles in stock turned out to be very labor intensive and error prone.
Read more
Zalando's modern logistic centers are equipped with automatic, highly accurate outbound scanners, which measure the weight and volume of every parcel leaving the warehouse on its journey to the customer. In this talk, I'll show that by using these parcel measurements we can accurately infer the weights of individual articles including confidence intervals, even in the common situation where most of the parcels contain more than one item. I'll explain the model we used, a Large Scale Bayesian Graphical Model, in detail and present weight estimation results in comparison to the manual weighing process.
ZalandoData Science Expert
