Personalized ranking for tag-based item recommendation system using tensor model

摘要

This research is a step forward in the study of generating item recommendations for the tag-based systems, in which two efficient tagging data interpretation schemes and four ranking methods are developed. The interpretation schemes apply ranking constraints to interpret the tagging data that allow a ranked representation and result in richer data. The ranking methods fall into the category of point-wise and list-wise based ranking approaches that consider the recommendation task as regression/classification and ranking respectively. This thesis, in particular, shows that tagging data interpretation schemes and learning-to-rank approaches play an important role in significantly improving the tag-based item recommendation quality.