Proposal of new item promotion among fashion e-commerce websites, using item clustering

Abstract

When we sale new items in e-commerce websites, we have a problem that we lack of customer data: access logs, buying history, etc.

Therefore, methods such as collaborative filtering, might not work in these cases.

In this study, we worked on item data clustering among Fashion e-commerce websites, and analogized which item cluster has a resemblance to the new item.

From there, we inquest and apply the promotion method: valid to the item cluster in the past, and discuss the results in the new case.