A Personalised Ranking Framework with Multiple Sampling Criteria for Venue Recommendation

Abstract

Recommending a ranked list of interesting venues to users based on their preferences has become a key functionality in LocationBased Social Networks (LBSNs) such as Yelp and Gowalla. Bayesian Personalised Ranking (BPR) is a popular pairwise recommendation technique that is used to generate the ranked list of venues of interest to a user, by leveraging the user's implicit feedback such as their check-ins as instances of positive feedback, while randomly sampling other venues as negative instances. To alleviate the sparsity that a ects the usefulness of recommendations by BPR for users with few check-ins, various approaches have been proposed in the literature to incorporate additional sources of information such as the social links between users, the textual content of comments, as well as the geographical location of the venues. However, such approaches can only readily leverage one source of additional information for negative sampling. Instead, we propose a novel Personalised Ranking Framework with Multiple sampling Criteria (PRFMC) that leverages both geographical in uence and social correlation to enhance the e ectiveness of BPR. In particular, we apply a multi-centre Gaussian model and a power-law distribution method, to capture geographical in uence and social correlation when sampling negative venues, respectively. Finally, we conduct comprehensive experiments using three large-scale datasets from the Yelp, Gowalla and Brightkite LBSNs. e experimental results demonstrate the e ectiveness of fusing both geographical in uence and social correlation in our proposed PRFMC framework and its superiority in comparison to BPR-based and other similar ranking approaches. Indeed, our PRFMC approach a ains a 37% improvement in MRR over a recently proposed approach that identi es negative venues only from social links.

A Contextual Attention Recurrent Architecture for Context-Aware Venue Recommendation

“…Such factorisation-based approaches assume that users who have visited similar venues share similar preferences, and hence are likely to visit similar venues in the future. Previous works on venue recommendation have shown that the contextual information associated with the users' observed feedback (time of the day, location) play an important role to enhance the effectiveness of CAVR as well as to alleviate the cold-start problem [6,7,22,30,32,34,36]. For example, Yao et al [30] extended the traditional MF-based approach by exploiting a high-order tensor instead of a traditional user-venue matrix to model multi-dimensional contextual information.…”

“…MF-based approaches typically aim to embed the users' and venues' preferences within latent factors, which are combined with a dot product operator to estimate the user's preference for a given venue. Approaches on MF typically encapsulate contextual information about the user, which can help to make effective recommendations for users with few historical checkins, known as the cold-start problem [22,30,32].…”

“…timestamp of a user's checkin and the geographical location of the checkin). Indeed, such contexts have been shown to play an important role in producing effective CAVR recommendations [6,22,30,32].…”

Unifying Explicit and Implicit Feedback for Rating Prediction and Ranking Recommendation Tasks

“…Indeed, BPR is a pairwise ranking function trained based on an underlying pointwise predictive function (usually matrix factorization, denoted as MF bpr ), which randomly samples unclicked items as negative instances. Since BPR was introduced, many enhancements have been proposed [10,12]. Indeed, while BPR is limited to only one type of implicit feedback, Loni et al [10] extended the original model to incorporate multiple types of implicit feedback.…”

Learning to Rank with Trust and Distrust in Recommender Systems

“…While several ranking models follow the pairwise learning strategy of BPR to exploit the selections of friends such as [17]- [19], they do not account for the selections of foes.…”