Improving Recommender Systems with contextual information
Recommender Systems are generally used to predict the preference that a user would have for a given item. These systems are widely deployed and we use them on a daily basis. Online stores such as Amazon use them to recommend related products to people whenever they are viewing a specific product. Content providers will track all the media content their users consume in order to recommend them more items. And even regular brick and mortar stores have been using loyalty cards for many years in order to track the buying habits of their customers.
In this project you will explore an existing framework that is used to implement and evaluate algorithms for Recommender Systems. More specific you will use Collaborative Filtering algorithms. These algorithms compare the feedback that users give for items to recommend them similar items. You will extend the existing algorithms so that besides a specific user rating, they can also use other contextual information to predict interesting items to users. As an example, Amazon is known to track the amount of time a user spends watching a specific product to determine how interested they are in the product. What we mean by contextual information can vary greatly depending on the application, think about the information from a users social network, his location...
The project will require some implementation work in Java, but you will spend a lot of your time performing simulations. For your experimental results to be valid it is important that you setup a good experiment.
Background & Literature
Recommender Systems Library: http://mloss.org/revision/view/1105/
Su, Xiaoyuan and Khoshgoftaar, Taghi M: A survey of collaborative filtering techniques