A user-centric pre-processing approach to improve the diversity of top-N recommendations while maintaining recommendation performance.
