The best recommendations
at the right time
for the right user.

Recommender systems can be found in many information systems. They are familiar from Amazon's product recommendations, Spotify's or last.fm's music recommendations or Netflix's film recommendations. The aim of a recommender system is to predict a user's interest in certain objects in order to recommend those objects to the user from a large number of objects for which the user's predicted interest is greatest. Recommender systems thus serve as a tool to cope with the flood of information.

Users may be interested in different objects at different times and in different situations. Everyone knows, for example, the inappropriate recommendations in web shops just because I recently bought a gift for my partner. A major challenge is to determine the current interest of a user in real time in order to be able to determine an optimal recommendation quantity at any point in time. One approach to this is the implementation of a recommender system with a data stream management system (DSMS).

As part of the data stream-based recommender systems project group, a concrete application scenario is to be implemented with the DSMS Odysseus, in which the current interest of all users in the objects of the software system is determined in real time in order to be able to recommend objects to users in a personalised and situation-adapted manner. Various data sources, such as social media data, usage data or context data (user location, etc.), are to be integrated and analysed for this purpose.

• Contact: Cornelius Ludmann
• Duration: 1 April 2015 to 31 March 2016 (SS 2015 + WS 2015/2016)