[1M3-GS-13] AI application: Social data and prediction

[1M3-GS-13-03] A Decision Model of Sales Period on List Price for Second-hand Fashion Items Based on Machine Learning Approach

〇Izumi Kuwata¹, Shimpei Kanazawa¹, Kenta Mikawa², Masayuki Goto¹, Hokuto Sasaki³ (1. Waseda University, 2. Shonan Institute of Technology, 3. ZOZO, Inc.)

Keywords:Second-hand Fashion items, Sales History Data, Proper Period, Estimating Sales Probability, Natural Gradient Boosting

In this study, we focus on a second-hand fashion EC site. The price down system this site introduce to avoid unsold items reduces the prices of items that have passed a certain period since their listing. Traditionally, the system has been made empirically. However, since it has a strong effect on profits, setting proper sales period on list price is important issues. On the other hand, we should take into account the sales period until the item is sold in order to set a proper sales period. Therefore, we proposed a method to decision the proper sales period on list price. In our proposed method, at first, we estimate the probability distribution of the sales period for each item. In addition, we repeat the estimation while changing the month of listing items. The effectiveness of our proposed method is clarified using real purchasing history data of this site.