JSAI2020

Presentation information

Interactive Session

[4Rin1] Interactive 2

Fri. Jun 12, 2020 9:00 AM - 10:40 AM Room R01 (jsai2020online-2-33)

[4Rin1-18] Consideration on Validity of Difficulty Estimated with VAE in Demand Forecasting

〇Yasuyuki Mitsui1, Yingsha Yang1, Kazuhiro Koike1 (1.Askul Corporation)

Keywords:Variational Auto-Encoder, Demand Forecasting

For retail business, highly accurate demand forecasting is very important. In e-commerce which treats particularly various kinds of items, however, it is difficult to improve accuracy of demand forecasting for all items, because variation of amount and frequency of orders for each item. In order to apply this problem and to discriminate the items which has difficulty of demand forecasting, we have proposed the method which estimates difficulty of forecasting by variational auto-encoder. In this paper, we estimate the difficulty of forecasting and forecast amount of shipments for each item, using performance data about real past shipments. By the result, we verify validity to estimate the difficulty of forecasting based on relation between the difficulty and the accuracy of forecasting and evaluate effectivity in real business.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.

Password