Web logs in e-commerce sites consist of user actions on items such as visiting an item description page, adding an item to a wishlist, and purchasing an item. Those items could be represented as nodes in a graph while viewing their relationships as edges according to the user actions. Based on the item graph, identifying items that attract users to purchase the target item could be practically used for supporting business decisions. To do this, we introduce a new task, called 'Purchase Influence Mining', that finds the top-k items (PIM-items) maximizing the estimated purchase influence from them to a target item. We solve this problem by modeling the purchase influence as the shortest path between item pair. According to the result, our approach more consistently finds the k PIM-items than the baseline.
|Title of host publication||WWW 2016 Companion - Proceedings of the 25th International Conference on World Wide Web|
|Publisher||Association for Computing Machinery, Inc|
|Number of pages||2|
|Publication status||Published - 2016 Apr 11|
|Event||25th International Conference on World Wide Web, WWW 2016 - Montreal, Canada|
Duration: 2016 May 11 → 2016 May 15
|Name||WWW 2016 Companion - Proceedings of the 25th International Conference on World Wide Web|
|Conference||25th International Conference on World Wide Web, WWW 2016|
|Period||16/5/11 → 16/5/15|
Bibliographical notePublisher Copyright:
© 2016 owner/author(s).
All Science Journal Classification (ASJC) codes
- Computer Networks and Communications