Research On Data Missing Problem In Recommender Systems

Recommender systems have been studied and deploy ed in various domains(e.g..e-commerce websites,online social networks.and online learning platforms).They become indispensable because they help manage information overload problems and provide cus-tomized recommendations b ased on users' preferences.At the meantime,the information explosion also brings challenges to researchers and developers,one of the most challenges is the missing data problem.Collab orative filtering(CF)is one of the most successful ap-proaches to recommendations.The core idea of CF is that users with similar tastes share similar rating distributions toward the same items.Popular methods combine CF with auxiliary information to make recommendationsUnder this setting,recommender system became a hot research area in both academia and industry.The data missing problem pose one of the most challenging problems in rec-ommender systems.Recommender systems are based on the historical interactions between users and items,but in reality only small part of users have interacted with limited number of items.which leads to the majority of data missing.Recommender systems can not learn sufficient good user preferences.This prob lem is also named"sparsity problem"in rcom-mendation area.Current recommendation techniques use side information to better learn user representations and item features.However,most existing studies have only focused on improving the accuracy of either item recommendations or tag recommendations.When viewing the recommendations from a unified perspective,we can utilize more coherent fea-tures and mine the relations between the two recommendations.On the other hand,most existing social recommender systems assumed users who are connected are more likely to have similar preferences.However,user characteristics change dynamically and why users interact in the social networks become diverse.It becomes easier to make friends today over the Internet and people make online friends for various reasons:e.g.,alumni,living in the same city,sharing similar interests.Online friends may not share similarities in preference.A more reasonable assumption is needed.Moreover,implicit feedback is widely used in collaborative filtering methods for recommendation.It is well known that implicit feedback contains a large number of values that are missing not at random(MNAR):and the missing data is a mixture of negative and unknown feedback.making it difficult to learn users' negative preferences.Recent studies modeled exposure.a latent missingness variable which indicates whether an item is exposed to a user,to give each missing entry a confidence of being negative feedback.However,these studies use static models and ignore the information in temporal dependencies among items,which seems to be an essential underlying factor to subsequent missingness.To solve the three issues related with the data missing problem,we propose three appro aches respectivey.The main contributions are as follows:· To fully utilize multi source contents,we propose a joint recommendation framework of users and items.Tag-based recommendation has become increas-ingly important in recent years owing to the popularization of social tagging systems.Related studies on this subject can be categorized into item recommendation and tag recommendation based on the objective of the recommendation.In social tagging sys-tems,tags not only behave as auxiliary information of items but also show the implicit preferences of users.Therefore,they can be used to improve prediction accuracy and provide explanations to recommended items.Items and tags have interrelation and mutual effects.By focusing only on either item recommendation or tag recommenda-tion.users may miss some information and can only achieve marginal gains.Fusing the two types of recommendations can improve the performance of both approaches.In this study,we propose an EXPLainable item-tag CO-REcommendation(EXPLORE)framework that jointly recommends items and the corresponding tags.Different from conventional recommendations,EXPLORE takes advantage of users' interests,item contents,and item tags.The experiments conducted on three real-world datasets demonstrate that EXPLORE outperforms the state-of-the-art methods.Importantly,the recommmended tags can provide explanations to the recommended items,thus making the recommendation results explainable.· To utilize noisy social data to alleviate the data missing problem,we pro-pose a collaborative filtering based recommendation framework using social exposure.Most existing social recommendation assume people share similar prefer-ences with their social friends.However,this assumption may not hold true due to various motivations of making online friends and dynamics of online social networks Inspired by recent causality based recommendations that first model user exposures towards items and then use these exposures to guide rating prediction.we utilize so-cial information to capture user exposures rather than user preferences.We assume that people get information of products from their online friends and they don't need to share similar preferences,which is more relaxed and closer to reality.In this paper we present a novel recommendation framework(named SERec)to integrate social ex-posure into collaborative filtering.We propose two methods to realize SERec,namely social regularization and social boosting,each with different intuition.Experiments on four real-world datasets demonstrate that our model outperforms the state-of-the-art models on top-N recommendations· To address the dynamic missing problem in implicit feedback,we propse a recommendation framework based on HMM and MF.Implicit feedback is widely used in collaborative filtering methods for recommendation.It is well known that implicit feedback contains a large number of values that are missing not at ran-dom(MNAR):and the missing data is a mixture of negative and unknown feedback,making it difficult to learn users' negative preferences.Recent studies modeled ex-posure,a latent missingness variable which indicates whether an item is exposed to a user,to give each missing entry a confidence of being negative feedback.However,these studies ignore the information in temporal dependencies among items,which seems to be an essential underlying factor to subsequent missingness.To model and exploit the dynamics of missingness,we propose a latent variable named"user intent"to govern the temporal changes of item missingness,and a hidden Markov model to represent such a process.The resulting framework captures the dynamic item miss-ingness and incorporate it into matrix factorization(MF)for recommendation.We also explore two types of constraints to achieve a more compact and interpretable rep-resentation of user intents.Experiments demonstrate the superiority of our method against state-of-the-art recommender systems.