Characterizing Users in a Classified Ad Network

  • Author / Creator
    Waqar, Muhammad
  • We study the problem of classifying users in a classified ad network and its applications in further analyzing the network. Specifically, we seek to classify Kijiji users into one of the two business and non-business categories. The problem is challenging due to the sparsity of the data about users, the vague separation of the two classes, and the highly imbalanced distribution of users between the classes. Our work utilizes the ad content to build a set of distinctive terms for each class (profile). Given the statistics on how an ad mentions terms from a class profile, the affinity of an ad (and subsequently a user) to a particular class is determined. Our experiments reveal that this is an effective strategy for classifying users, outperforming various baselines. We study the impact of profile size on the classification task and observe that using longer class profiles may not be helpful. Moreover, in the absence of labeled training data, we show that a simple bootstrapping technique with only a few n-grams as a seed set can give nearly good results in terms of F-measure. We also study the same problem from a different angle: collective behavior of a user in posting ads. Using features associated with such behavior, we identify four distinct usage patterns for the users of the Kijiji network and study the association of business and non-business users with these patterns. Our experiments reveal that a sizeable number of members from both user groups validly manifest all the patterns, due to which the aforementioned features are inadequate for the classification task. Finally, using the results of user classification, we analyze the Kijiji network from various aspects. Our results, for example, indicate that businesses are more amenable to post consistently in a particular set of categories than non-business users and that the popularity of different categories for both the user groups exhibits various seasonal trends.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
    • Department of Computing Science
  • Supervisor / co-supervisor and their department(s)
    • Rafiei, Davood (Computing Science)
  • Examining committee members and their departments
    • Moore, Sarah (Business)
    • Rafiei, Davood (Computing Science)
    • Wong, Kenny (Computing Science)
    • Barbosa, Denilson (Computing Science)