Recommending ephemeral items at web scale


  • Generative Clustering Model
    • Objective
      • To maximize the total intra-cluster coherence
    • Use-cases
      • Naive Bayes for ranking
      • (Similar product) Inferring latent product for unseen one
  • Evaluate using offline and online
    • Offline evaluation
      • The area under the click-view ROC curve (AOC)
        • Plots the click recall vs view recall from the testing examples ranked in descending order
      • Relative CTR lift over a baseline predictor at a certain recall level of view
    • Online evaluation
      • CTR
      • BTR (bid-through rate)
      • PTR (purchase-through rate
      • GMB (gross merchandising bought)


  • 3-5 folds improvement in CTR and Purchase-through rate


Y. Chen and J. F. Canny, “Recommending ephemeral items at web scale,” in Proceedings of the 34th international ACM SIGIR conference on Research and development in Information - SIGIR ’11, Beijing, China, 2011, p. 1013. doi: 10.1145/2009916.2010051.