Incorporating Contextual Information into Personalized Recommendations
Speaker: Shanle Ma, ITEE
When: 2007-02-06 14:00:00
Venue: 78-622
Host: Dr Xue Li
Abstract:The tremendous growth of information on the Internet has been over
our ability to process. Recommender systems, which filter for useful
information and generate recommendations, were introduced to help
users overcome the information overload problem and has been widely
applied in an ever-increasing number of e-commerce websites.
Collaborative filtering and content-based recommendation methods are
two major approaches used in recommender systems. Collaborative
filtering method predicts items which a user might like by using a
database about other users past preferences. Content-based method
analyze the content of the objects to generate a representation of
the user's interests, and then compare the similarity of item
descriptions. These two methods have some drawbacks in dealing with
situations such as sparse data and cold start problems. Recently,
hybrid methods combining collaborative filtering and content-based
methods were proposed to overcome these limitations.
However, a highly effective and personalized recommender system may
still face a new challenge on incorporating contextual
information. In this case, recommender system may give individual
user tailored recommendations based not only on traditional
user-item matrix but also on their contextual information such as
user's location, companion,demand and so on . For example, user's
preference may change with their location changing. Another example
is that user's demand may decline over time or cycle with
time. Unfortunately, current available Recommender systems do not
consider this important factor. In our work, we proposed a novel
hybrid recommender system to overcome the individual user's demand
declining by embedding the time-sensitive functions into the
recommendation process. We also proposed a method based on user's
cycling demand.
Our experiments demonstrate a better performance than that of the
collaborative filtering approaches considering interests drift and
those of the combined approaches without considering demand
declining.
This report will discuss the current performance of recommender
systems and their challenges. We also present some possible
approaches to improve the performance of recommendations.
Biography:(biography unavailable)
Type: MPhil confirmation
Contact:Dr Xue Li, seminar host (xueli@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)