• Offline process, generates long-term cluster based on product’s title
  • e.g.
    • c1 = {nike, air-max, white, gray, running}
      • Nike white gray 7 mesh Air Max -> c1
      • WOMENS 10.5 Nike Air Max Navi. Gray pink white running shoes -> c1
• Maintains similarity “importance” weight for each cluster by matching keywords against known entities (brand=Nike)
• Associate statistics with clusters based on cluster dictionary.

Results

• 38.18% increase in CTR
• 89.6% increase in saving to “watch-list”

Reference

J. Katukuri, T. Konik, R. Mukherjee, and S. Kolay, “Recommending similar items in large-scale online marketplaces,” in 2014 IEEE International Conference on Big Data (Big Data), Washington, DC, USA, Oct. 2014, pp. 868–876. doi: 10.1109/BigData.2014.7004317.

#recommender-system #recommender-system/ecommerce