k-Nearest Neighbor (k-NN) and other user-based collaborative filtering (CF) algorithms have gained popularity because of the simplicity of their algorithms and performance. As the performance of such algorithms largely depends on neighborhood selection, it is important to select the most suitable neighborhood for each active user. Previous user-based CF simply relies on similar users or common experts in this regard; however, because users have different tastes as well as different expectations for expert advice, similar users or common experts may not always be the best neighborhood for CF. In search of a more suitable neighborhood, so-called personalized experts develop personalized expert features. Through experimentation, we show that personalized experts are different from similar users, common experts, or similar common experts. The personalized, expert-based CF algorithm outperforms k-NN and other user-based CF algorithms.