Serendipitous Personalized Ranking for Top-N Recommendation

Abstract

Serendipitous recommendation has benefitted both e-retailers and users. It tends to suggest items which are both unexpected and useful to users. These items are not only profitable to the retailers but also surprisingly suitable to consumers' tastes. However, due to the imbalance in observed data for popular and tail items, existing collaborative filtering methods fail to give satisfactory serendipitous recommendations. To solve this problem, we propose a simple and effective method, called serendipitous personalized ranking. The experimental results demonstrate that our method significantly improves both accuracy and serendipity for top-N recommendation compared to traditional personalized ranking methods in various settings.

A Serendipity-Oriented Greedy Algorithm for Recommendations

“…A change of serendipity might affect other properties of a recommender system. To demonstrate the dependence of different properties and features of the baselines, we employed evaluation metrics to measure four properties of recommender systems: (1) accuracy, as it is a common property (Kotkov et al, 2016b), (2) serendipity, as SPR, Zhengs, SOG and SOGBasic are designed to improve this property (Lu et al, ;Zheng et al, 2015), (3) diversity, as this is one of the objectives of TD, SOG and SOGBasic (Ziegler et al, 2005) and (4) novelty, as SPR, Zhengs, SOG and SOGBasic are designed to improve this property (Lu et al, ;Zheng et al, 2015).…”

“…We selected the 100 most popular items for PM following one of the most common strategies (Zheng et al, 2015;Lu et al, ). Items relevant to user u are represented by REL u , REL u = {i ∈ TestSet u |r ui > θ}, where TestSet u is a ground truth for user u, while θ is the threshold rating, which in our experiments equals to 3.…”

“…In different scenarios, ways to express preference might vary. For example, we might regard a movie relevant to a user if the user gave it more than 3 stars out of 5 (Zheng et al, 2015;Lu et al, ), while we might regard a song relevant to a user if the user listened to it more than twice. The system is aware that a particular item is relevant to a user if the user rates the item, and unaware of this relevance otherwise.…”

“…State-of-the-art serendipity-oriented recommendation algorithms are barely compared with one another and often employ different serendipity metrics and definitions of the concept, as there is no agreement on the definition of serendipity in recommender systems (Zhang et al, 2012;Lu et al, ;Kotkov et al, 2016b).…”

LambdaFM

“…In fact, it has been recognized that item popularity distribution in most real-world recommendation datasets has a heavy tail [17,13], following an approximate power-law or exponential distribution 4 . Accordingly, most non-positive items drawn by uniform sampling in Algorithm 1 are unpopular due to the long tail distribution, and thus contribute less to the desired loss function.…”