Master of Science (MS)
Everyday, millions of people interact with online services that adopt recommender systems, such as personalized movie, news and product recommendation services. Research has shown that the demographic attributes of users such as age and gender can further improve the performance of recommender systems and can be very useful for many other applications such as marketing and social studies. However, users do not always provide those details in their online profiles due to privacy concern. On the other hand, user interactions such as ratings in recommender systems may provide an alternative way to infer demographic information. Most existing approaches can infer user demographics based on sufficient interaction history but could fail for users with few ratings. In this thesis, we study the association between users demographic information and their ratings, and explore the tradeoff between user privacy and the utility of personalization. In particular, we present a novel multi-task preference elicitation method, with which a recommender system asks a new user to rate selected items adaptively and infers the demographics rapidly via a few interactions. Experimental results on real-world datasets demonstrate the performance of the proposed method in terms of the accuracy of both demographics inference and rating prediction.
Document Availability at the Time of Submission
Secure the entire work for patent and/or proprietary purposes for a period of one year. Student has submitted appropriate documentation which states: During this period the copyright owner also agrees not to exercise her/his ownership rights, including public use in works, without prior authorization from LSU. At the end of the one year period, either we or LSU may request an automatic extension for one additional year. At the end of the one year secure period (or its extension, if such is requested), the work will be released for access worldwide.
Li, Changbin, "A Study on User Demographic Inference Via Ratings in Recommender Systems" (2017). LSU Master's Theses. 4466.
Available for download on Saturday, February 23, 2019