Unifying Nearest Neighbors Collaborative Filtering

“We study collaborative filtering for applications in which there exists for every user a set of items about which the user has given binary, positive-only feedback (one-class collaborative filtering). Take for example an on-line store that knows all past purchases of every customer. An important class of algorithms for one-class collaborative filtering are the nearest neighbors algorithms, typically divided into user-based and item-based algorithms. We introduce a reformulation that unifies user- and item-based nearest neighbors algorithms and use this reformulation to propose a novel algorithm that incorporates the best of both worlds and outperforms state-of-the-art algorithms. Additionally, we propose a method for naturally explaining the recommendations made by our algorithm and show that this method is also applicable to existing user-based nearest neighbors methods.”

“We introduce a reformulation of these algorithms that unifies their existing formulations. From this reformulation, it becomes clear that the existing user- and item-based algorithms unnecessarily discard important parts of the available information. Therefore, we propose a novel algorithm that combines both user- and item-based information. Hence, this algorithm is neither user-based nor item-based, but nearest-neighbors-based. Our experiments show that our algorithm not only outperforms the individual nearest neighbors algorithms but also state-of-the-art matrix factorization algorithms.”

“Intuitively, a more popular item is considered less informative for determining a taste, since this item is more likely preferred by diverse users. Similarly, a user that prefers many items is considered less informative for determining a taste, since this user's preferences are more likely to cover diverse items.”