Recommender Response to User Profile Diversity and Popularity Bias
Recommender systems are commonly evaluated to understand the effectiveness of their algorithms. Diversity and novelty of the recommender systems have been in consideration while evaluating the systems in addition to accuracy and prediction metrics in order to provide better recommendations. Different evaluation metrics that are related to diversity and novelty have been discussed in some of the previous works. This work provides a comprehensive study and analysis of the recommender algorithms and its relationship to the user’s bias in terms of popularity and diversity. This kind of analysis helps us to understand if the core algorithms personalize the recommendations based on the users’ bias. We performed offline experiments using the MovieLens data and analyzed the correlation between the user profile and the recommender profile for both diversity and popularity bias using different metrics. Finally, we report the analysis observations and study how it complements the previous work done.
Recommender systems, Recommender
Channamsetty, S. (2016). <i>Recommender response to user profile diversity and popularity bias</i> (Unpublished thesis). Texas State University, San Marcos, Texas.