No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Using a trust network to improve top-N recommendation
Abstract
Top-N item recommendation is one of the important tasks of rec- ommenders. Collaborative filtering is the most popular appr oach to building recommender systems which can predict ratings for a given user and item. Collaborative filtering can be extended for top- N recommendation, but this approach does not work accurately for cold start users that have rated only a very small number of items. In this paper we propose novel methods exploiting a trust network to improve the quality of top-N recommendation. The first met hod performs a random walk on the trust network, considering the sim- ilarity of users in its termination condition. The second me thod combines the collaborative filtering and trust-based appro ach. Our experimental evaluation on the Epinions dataset demonstrates that approaches using a trust network clearly outperform the collabora- tive filtering approach in terms of recall, in particular for cold start users.
... However, the popular products for other customers may not be interesting for a specific user. Recently, some random-walk-based approaches [15,16,36,47] have been proposed to employ the random-walk process on the rating dataset to model the preference propagation among users and thus alleviate the cold start problem. Serval social-network-based methods [17,27,29,46] have been proposed to integrate the rating matrix and the auxiliary social networks into unified recommender systems to reduce negative effect of data sparsity and thus improve the prediction accuracy. ...
... The common rationale behind all of the social recommendation solutions is that a user's taste is similar to his/her neighbor's taste, be it a social friend or a user with similar rating behavior. Recently, IRCD-CCS and IRCD-ICS were proposed to integrate a deep neural network, SADE, for extracting the content features of items, and a recent CF model, timeSVD++, for utilizing temporal dynamics of user preferences and item features, to solve the complete cold start and the incomplete cold start problems in the recommendation [43]; (2) random-walk-based approaches [15,16,36,47] employ the random-walk process on the rating dataset to model the preference propagation among users. The preference similarity between users is learnt to reduce the problem of cold start with new users; and (3) other recommendation solutions [4,8,11,18,48,51,53]. ...
... T inf luence_pr opaдation is bounded by (16). ...
The advent of social networks and activity networks affords us an opportunity of utilizing explicit social information and activity information to improve the quality of recommendation in the presence of data sparsity. In this article, we present a social-influence-based collaborative filtering (SICF) framework over heterogeneous information networks with three unique features. First, we integrate different types of entities, links, attributes, and activities from rating networks, social networks, and activity networks into a unified social-influence-based collaborative filtering model through the intra-network and inter-network social influence. Second, we propose three social-influence propagation models to capture three kinds of information propagation within heterogeneous information networks: user-based influence propagation on user rating networks, item-based influence propagation on user-rating activity networks, and term-based influence propagation on user-review activity networks, respectively. We compute three kinds of social-influence-based user similarity scores based on three social-influence propagation models, respectively. Third, a unified social-influence-based CF prediction model is proposed to infer rating tastes by incorporating three kinds of social-influence-based similarity measures with different weighting factors. We design a weight-learning algorithm, SICF, to refine the prediction result by quantifying the contribution of each kind of information propagation to make a good balance between prediction accuracy and data sparsity. Extensive evaluation on real datasets demonstrates that SICF outperforms existing representative collaborative filtering methods.
... Jamali & Ester [52,53] Graph Structure-based Recommendations: The Random Walk of a trusted network and item-based collaborative filtering with weighted hybridization, socalled 'TrustWalker' RMSE, coverage, F-measure, and cold-start user problem Jamali & Ester [54] Matrix Factorization: Matrix Factorization combined with users' trust-based networks -SocialMF. In particular, latent feature vectors of users were weighted by average ratings of users' direct trusted social links. ...
... As a way to integrate two different types of recommendations together -trustbased recommendation and item-based collaborative filtering -Jamali and Ester used a random walk model, so-called 'TrustWalker' [52,53] in the context of Epinions.com data set. ...
... This process continues recursively until rating values of the candidate is found among a target user's direct and indirect trusted links. However, in order to prevent walking too far in the trust-based network, if directly or indirectly trusted user rated an item which is quite similar to the candidate item and the similarity weighted by the distance is above a certain threshold, the algorithm stops the walking and returned the ratings of similar item [52,53]. ...
Springer International Publishing AG, part of Springer Nature 2018. The goal of this chapter is to give an overview of recent works on the development of social link-based recommender systems and to offer insights on related issues, as well as future directions for research. Among several kinds of social recommendations, this chapter focuses on recommendations, which are based on users’ self-defined (i.e., explicit) social links and suggest items, rather than people of interest. The chapter starts by reviewing the needs for social link-based recommendations and studies that explain the viability of social networks as useful information sources. Following that, the core part of the chapter dissects and examines modern research on social link-based recommendations along several dimensions. It concludes with a discussion of several important issues and future directions for social link-based recommendation research.
... While graph-based recommendation approaches have been originally explored for traditional recommendations, they become especially popular in the area of social link-based recommendation because social links could be most naturally represented as a social graph. Among the projects reviewed in Table 3, seven studies used various graph-based recommendation algorithms (Bellogín et al. [8], Deng et al. [32], Jamali & Ester [54,55], Konstas et al. [68], Wang et al. [134], Yuan et al. [146]). Bellogín et al. [8] suggested a social recommendation approach based on users' friendship network using Breadth-First Search algorithm. ...
... As a way to integrate two different types of recommendations together -trustbased recommendation and item-based collaborative filtering -Jamali and Ester used an RW model TrustWalker [54,55] in the context of Epinions.com data set. ...
... This process continues recursively until rating values of the candidate is found among a target user's direct and indirect trusted links. However, in order to prevent walking too far in the trustbased network, if directly or indirectly trusted user rated an item which is quite similar to the candidate item and the similarity weighted by the distance is above a certain threshold, the algorithm stops the walking and returned the ratings of similar item [54,55]. ...
The goal of this chapter is to give an overview of recent works on the development of social link-based recommender systems and to offer insights on related issues, as well as future directions for research. Among several kinds of social recommendations, this chapter focuses on recommendations, which are based on users’ self-defined (i.e., explicit) social links and suggest items, rather than people of interest. The chapter starts by reviewing the needs for social link-based recommendations and studies that explain the viability of social networks as useful information sources. Following that, the core part of the chapter dissects and examines modern research on social link-based recommendations along several dimensions. It concludes with a discussion of several important issues and future directions for social link-based recommendation research.
... The former mainly employ memory-based collaborative filtering methods-they search the trust networks to obtain the neighbors and then make recommendations based on those trusted neighbors [17]. For example, Jamali and Ester [12] combine TrustWalker [11] with neighborhood collaborative filtering. They first use random walks to get the user representation from the trust network and then perform a probabilistic strategy for selecting items to give recommendation. ...
... Specifically, the model propagates trust information over the social trust network to estimate the weight for the trust link that can be used in place of the user similarity weight. Jamali and Ester [12] combine TrustWalker [11] with neighborhood collaborative filtering. They first use random walks to get the user representation from the trust network and then perform a probabilistic strategy for selecting items to give Neural Computing and Applications recommendation. ...
Recommender systems face longstanding challenges in gaining users’ trust due to the unreliable information caused by profile injection or human misbehavior. Traditional solutions to those challenges focus on leveraging users’ social relationships for inferring the user preference, i.e., recommending items according to the preference by user’s trusted friends; or adding random noise to the input to improve the robustness of the recommender systems. However, such approaches cannot defend the real-world noises like fake ratings. The recommender model is generally built upon all the user-item interactions, which incorporates the information from fake ratings or spammer groups, that neglects the reliability of the ratings. To address the above challenges, we propose an adversarial training approach in this work. In details, our approach includes two components: a predictor that infers the user preference; and a discriminator that enforces cohort rating patterns. In particular, the predictor applies an encoder-decoder structure to learn the shared latent information from sparse users’ ratings and trust relationships; the discriminator enforces the predictor to provide ratings as coherent with the cohort rating patterns. Our extensive experiments on three real-world datasets show the advantages of our approach over several competitive baselines.
... presence of data sparsity [34][35][36][37][38][39][40][41][42][43][44]. A recent study reports that social information can not always result in substantial performance gains of CF prediction, and sometimes may lead to biased prediction result [45]. ...
... The social network enhanced CF methods can be classified into two categories: (1) matrix factorization or network embedding based methods [34][35][36][37][38][39][40][41], which integrate the user friendship or user trust relationship in a social network into the basic matrix factorization or network embedding models by factorizing the social network or capturing the dependencies between feature vectors of users and their social neighbors; and (2) neighborhood based approaches [42][43][44], which combine the basic memory-based solutions and the social network based approaches to improve the recommendation performance. ...
This paper presents a multi-label classification based CF framework, MLCF, which improves the quality of recommendation in the presence of data sparsity by learning over a heterogeneous information network consisting of a rating bipartite graph, a user graph and an item graph. MLCF is novel by three unique features. First, we explore the latent correlations among users and items w.r.t. a given set of K semantic categories beyond user-item ratings by employing multi-label clustering of items, and multi-label classification of users and rating-based similarities on the heterogeneous network. Second, based on the user/item/similarity multi-label clustering/classification, we propose a fine-grained multi-label classification based rating similarity measure to capture the class-specific relationships between users by introducing a novel concept of vertex-edge homophily. Third but not the least, we propose to integrate two kinds of multi-label classification based CF models focusing on rating and social information into a unified prediction model.
... Therefore, when recommending items for a user, it is more sensible to take into account both trusters and trustees influence simultaneously. Although existing item ranking literatures [1]- [8] have proved that trusted users will influence items' ranking, they neglect the influence of trusters as well as its contribution to ranking generation process. Based on this fact, in this work, we take the view of "dual roles influence" to explain the decision making of item selection by individual users and leverage it to enhance the existing trust-based ranking models. ...
... Then, we survey some representative trust-enhanced item ranking approaches. Jamali and Ester extend TrustWalker to perform Top-K recommendation namely Trust-CF [1]. Different from TrustWalker, Trust-CF weights trusted users by their correlation with the source user instead of equally treating for all trusted users. ...
Ranking items to users is a typical recommendation task, which evaluates users’ preferences for certain items over others. Easy access to social networks has motivated researchers to incorporating trust information for recommendation. In this paper, aiming at offering fundamental support to the trust-based research for item recommendation, we conduct an in-depth analysis on Epinions, Ciao, and FilmTrust data sets. We find that a user’s selection of an item is influenced not only by her trustees but also by her trusters. We leverage this “dual roles influence” to derive two more accurate matrix factorization (MF)-based ranking models, namely,
BPRDR
and
FSDR
, respectively. In more detail, the first
BPRDR
model performs three pairwise preferences comparisons under the Bayesian personal ranking framework, considering the dual roles influence in its ranking assumptions. The second
FSDR
is an improved factored similarity model as it incorporates dual roles influence to contribute its ranking scores. Extensive experiments on three data sets show that it is essential to consider the dual roles influence when generating top-
$K$
item recommendation.
... Dans l'approche à base de confiance, l'impact du démarrage à froid est réduit par rapport au FC Bhattacharjee2004, Jamali andEster2009]. Il suffit pour un utilisateur du SRC de noter un seul utilisateur, pour créer un lien entre les deux utilisateurs. ...
... L'approche à base de confiance s'appuie sur les notes mutuelles entre utilisateurs, cela est plus efficace pour remplir la matrice de similarité [Jamali andEster2009, Lee andBrusilovsky2009]. De plus, la propagation de confiance sert aussi le même but. ...
La divergence comportementale des utilisateurs sur le web résulte un problème de fluctuation de performance chez les systèmes de recommandation (SR) qui exploitent ce comportement pour recommander aux utilisateurs des items qu’ils vont apprécier. Ce problème est observé dans l’approche de filtrage collaboratif (FC) qui exploite les notes attribuées par les utilisateurs aux items, et l’approche à base de confiance (SRC) qui exploite les notes de confiance que les utilisateurs attribuent l’un à l’autre. Nous proposons une approche hybride qui augmente le nombre d'utilisateurs bénéficiant de la recommandation, sans perte significative de précision. Par la suite, nous identifions plusieurs caractéristiques comportementales qui permettent de constituer un profil comportemental de l’utilisateur. Ce qui nous permet de classifier les utilisateurs selon leur comportement commun, et d’observer la performance de chaque approche par classe. Par la suite, nous focalisons sur les SRC. Le concept de confiance a été abordé dans plusieurs disciplines. Il n'existe pas véritablement de consensus sur sa définition. Cependant, toutes s'accordent sur son effet positif. La logique subjective (LS) fournit une plateforme flexible pour modéliser la confiance. Nous l’utilisons pour proposer et comparer trois modèles de confiance, dont l’objectif est de prédire à un utilisateur source s’il peut faire confiance à un utilisateur cible. La recommandation peut s’appuyer sur l’expérience personnelle de la source (modèle local), un système de bouche à oreille (modèle collectif), ou encore la réputation du cible (modèle global). Nous comparons ces trois modèles aux termes de la précision, la complexité, et la robustesse face aux attaques malicieuses
... Collaborative filtering effectively bridges the user space and the item space, but does not require the content analysis of the item, rendering it a versatile and widely used algorithm for recommender system studies. For example, Jamali and Ester (2009) developed an approach to incorporating a social network into a top-N recommender system based on nearest neighbors. Liu et al. (2013) proposed a Bayesian probabilistic matrix factorization algorithm that was applied to trust-aware recommender systems for large datasets. ...
Motivated by the connections between collaborative filtering and network clustering, we consider a network-based approach to improving rating prediction in recommender systems. We propose a novel Bipartite Mixed-Membership Stochastic Block Model ($\mathrm{BM}^2$) with a conjugate prior from the exponential family. We derive the analytical expression of the model and introduce a variational Bayesian expectation-maximization algorithm, which is computationally feasible for approximating the untractable posterior distribution. We carry out extensive simulations to show that $\mathrm{BM}^2$ provides more accurate inference than standard SBM with the emergence of outliers. Finally, we apply the proposed model to a MovieLens dataset and find that it outperforms other competing methods for collaborative filtering.
... Leave-One-Out Memorization Task. In this task, we adapt the widely adopted leave-one-out protocol [46,47,58,65] to evaluate memorization. Different from the original protocol, we do not hold out a test item, but pick an item that was seen during training. ...
In recent years, neural models have been repeatedly touted to exhibit state-of-the-art performance in recommendation. Nevertheless, multiple recent studies have revealed that the reported state-of-the-art results of many neural recommendation models cannot be reliably replicated. A primary reason is that existing evaluations are performed under various inconsistent protocols. Correspondingly, these replicability issues make it difficult to understand how much benefit we can actually gain from these neural models. It then becomes clear that a fair and comprehensive performance comparison between traditional and neural models is needed. Motivated by these issues, we perform a large-scale, systematic study to compare recent neural recommendation models against traditional ones in top-n recommendation from implicit data. We propose a set of evaluation strategies for measuring memorization performance, generalization performance, and subgroup-specific performance of recommendation models. We conduct extensive experiments with 13 popular recommendation models (including two neural models and 11 traditional ones as baselines) on nine commonly used datasets. Our experiments demonstrate that even with extensive hyper-parameter searches, neural models do not dominate traditional models in all aspects, e.g., they fare worse in terms of average HitRate. We further find that there are areas where neural models seem to outperform non-neural models, for example, in recommendation diversity and robustness between different subgroups of users and items. Our work illuminates the relative advantages and disadvantages of neural models in recommendation and is therefore an important step towards building better recommender systems.
... Metric. We use two widely used metrics Recall@K [19] and NDCG@K [20] to test the performance of all experiments. ...
Bundle recommendation is an emerging research direction in the recommender system with the focus on recommending customized bundles of items for users. Although Graph Neural Networks (GNNs) have been applied in this problem and achieve superior performance, existing methods underexplore the graph-level GNN methods, which exhibit great potential in traditional recommender system. Furthermore, they usually lack the transferability from one domain with sufficient supervision to another domain which might suffer from the label scarcity issue. In this work, we propose a subgraph-based Graph Neural Network model, SUGER, for bundle recommendation to handle these limitations. SUGER generates heterogeneous subgraphs around the user-bundle pairs, and then maps those subgraphs to the users' preference predictions via neural relational graph propagation. Experimental results show that SUGER significantly outperforms the state-of-the-art baselines in both the basic and the transfer bundle recommendation problems.
... The memory-based methods deduce ratings of a targeted user via trust propagation based on ratings of its friends [27]. For example, Jamali and Ester [28] combine TrustWalker [29] with neighborhood collaborative filtering. They first run random walks on the trust network and then perform a probabilistic item selection strategy to generate recommendations. ...
A significant remaining challenge for existing recommender systems is that users may not trust recommender systems for either inaccurate recommendation or lack of explanation. Thus, it becomes critical to embrace a trustworthy recommender system. This survey provides a systematic summary of three categories of trust issues in recommender systems: social-aware recommender systems, which leverage users’ social trust relationships; robust recommender systems, which filter untruthful information, noises and enhance attack resistance; and explainable recommender systems, which provide explanations of the recommended items. We focus on the work based on deep learning techniques, which is an emerging area in the recommendation research.
... Jamali and Ester developed a traditional memory-based social-aware RS approach by combining TrustWalker and neighbourhood collaborative filtering [28]. To create recommendations, they first do random walks over the social-trust network and then use a probabilistic item choosing strategy. ...
Conventionally, cold-start limitations are managed by leveraging side information such as social-trust relationships. However, the relationships between users in social networks are complex, uncertain, and sparse. Therefore, it is necessary to extract beneficial social connections to make the recommendation models cold-start resistant. Towards this end, we propose a novel recommendation model called Variational Cold-start Resistant Recommendation (CORE-VAE). More concretely, we employ a social-aware similarity function and a graph convolutional network (GCN) to generate robust social-aware user representations that account for the complexities, uncertainties, and sparse nature of the social-trust network. Subsequently, these powerful social-aware representations aid us in producing cold-start resistant rating vectors for all users. To explore the rich user rating information, we propose an expressive variational autoencoder (VAE) model. Unlike earlier VAE-based CF models, CORE-VAE utilizes a novel prior distribution and a well-designed skip-generative network to alleviate the posterior collapse issue considerably. Besides, CORE-VAE can also capture the latent space’s uncertainty and ensure that observations and their accompanying latent variables have high mutual information. Overall, these novel techniques dramatically help produce better latent representations for generating more accurate recommendations. We show that CORE-VAE outperforms numerous competitive baseline models on real-world datasets through comprehensive empirical evaluation and analysis.
... Authors have used the social networks data in order to produce best recommendation system for the user. In [57], the authors have used the information of the social network in order to improve the Top-k recommendations and has been improvised using the Matrix Factorization and Nearest Neighbor algorithms [150]. In [138], the authors have proposed a tourist recommendation system named as GuideMe which integrates the social networks and the unique set of options which are given in the applications. ...
A Recommendation System (RS) is an intelligent computer based system which provide valuable suggestions to the user and are used in several domains. Social media platforms are the most common internet applications due to the large number of users. The numerous posts, likes, etc. have accrued on social media platforms and can be used in variety of recommendation systems. In this work, the primary focus is the tourism domain, where RS serves as a valuable tool for the tourist to plan his trip. Traditional RS systems only cater to the needs of the tourist by examining few factors. However, there are a large range of factors such as environment factors , actual geo-coordinates, trip destination, preferences of the user etc. that need to be taken into account in order to make a foolproof recommendation to the tourists. Tourism Recommendation Systems (TRS) provide suggestions to the tourists to identify the most suited transport (flight, train, etc.), accommodations, museums, special interest places and other items which are required for the trip. Several techniques are used and a thorough study of various techniques of traditional RS and TRS techniques have been done which are specially designed for tourism domain. Various Artificial Intelligence (AI) techniques have been highlighted which are used to solve the tourist recommendation problem. Also, future research direction has been suggested which would improvise the Quality of Service (QoS) of the RS in tourism industry.
... Microblog contents has general feedback on personalization. Distributing content on the microblog creates problems for suggestions [36,37]. e tags that users tag for content use user-tag matrix to solve the distribution problem, while user-user matrix may precisely cluster users and make appropriate recommendations [36]. ...
Nowadays, the recommendation is an important task in the decision-making process about the selection of items especially when item space is large, diverse, and constantly updating. As a challenge in the recent systems, the preference and interest of users change over time, and existing recommender systems do not evolve optimal clustering with sufficient accuracy over time. Moreover, the behavior history of the users is determined by their neighbours. The purpose of the time parameter for this system is to extend the time-based priority. This paper has been carried out a time-aware recommender systems based on memetic evolutionary clustering algorithm called RecMem for recommendations. In this system, clusters that evolve over time using the memetic evolutionary algorithm and extract the best clusters at every timestamp, and improve the memetic algorithm using the chaos criterion. The system provides appropriate suggestions to the user based on optimum clustering. The system uses optimal evolutionary clustering using item attributes for the cold-start item problem and demographic information for the cold start user problem. The results show that the proposed method has an accuracy of approximately 0.95, which is more effective than existing systems.
... Microblog contents has general feedback on personalization. Distributing content on the microblog creates problems for suggestions [36,37]. e tags that users tag for content use user-tag matrix to solve the distribution problem, while user-user matrix may precisely cluster users and make appropriate recommendations [36]. ...
Nowadays, the recommendation is an important task in the decision-making process about the selection of items especially when item space is large, diverse, and constantly updating. As a challenge in the recent systems, the preference and interest of users change over time, and existing recommender systems do not evolve optimal clustering with sufficient accuracy over time. Moreover, the behavior history of the users is determined by their neighbours. e purpose of the time parameter for this system is to extend the time-based priority. is paper has been carried out a time-aware recommender systems based on memetic evolutionary clustering algorithm called RecMem for recommendations. In this system, clusters that evolve over time using the memetic evolutionary algorithm and extract the best clusters at every timestamp, and improve the memetic algorithm using the chaos criterion. e system provides appropriate suggestions to the user based on optimum clustering. e system uses optimal evolutionary clustering using item attributes for the cold-start item problem and demographic information for the cold start user problem. e results show that the proposed method has an accuracy of approximately 0.95, which is more effective than existing systems.
... Les informations sur la confiance sont intéressantes pour améliorer les recommandations, en particulier pour le démarrage à froid des utilisateurs (utilisateurs pour lesquels l'historique des actions passées est très limité). Comme mentionné dans le chapitre 4, certains systèmes de recommandation incorporent des informations explicites sur la confiance entre les utilisateurs [68,52,108]. Cependant, ces informations explicites étant rarement disponibles, plusieurs systèmes de recommandation utilisent la confiance implicite [110,81]. ...
La recommandation des produits appropriés aux clients est cruciale dans de nombreuses plateformes de e-commerce qui proposent un grand nombre de produits. Les systèmes de recommandation sont une solution favorite pour la réalisation de cette tâche. La majorité des recherches de ce domaine reposent sur des notes explicites que les utilisateurs attribuent aux produits, alors que la plupart du temps ces notes ne sont pas disponibles en quantité suffisante. Il est donc important que les systèmes de recommandation utilisent les données implicites que sont des flots de liens représentant les relations entre les utilisateurs et les produits, c'est-à-dire l'historique de navigation, des achats et de streaming. C'est ce type de données implicites que nous exploitons. Une approche populaire des systèmes de recommandation consiste, pour un entier N donné, à proposer les N produits les plus pertinents pour chaque utilisateur : on parle de recommandation top-N. Pour ce faire, bon nombre de travaux reposent sur des informations telles que les caractéristiques des produits, les goûts et préférences antérieurs des utilisateurs et les relations de confiance entre ces derniers. Cependant, ces systèmes n'utilisent qu'un ou deux types d'information simultanément, ce qui peut limiter leurs performances car l'intérêt qu'un utilisateur a pour un produit peut à la fois dépendre de plus de deux types d'information. Pour remédier à cette limite, nous faisons trois propositions dans le cadre des graphes de recommandation. La première est une extension du Session-based Temporal Graph (STG) introduit par Xiang et al., et qui est un graphe dynamique combinant les préférences à long et à court terme des utilisateurs, ce qui permet de mieux capturer la dynamique des préférences de ces derniers. STG ne tient pas compte des caractéristiques des produits et ne fait aucune différence de poids entre les arêtes les plus récentes et les arêtes les plus anciennes. Le nouveau graphe proposé, Time-weight content-based STG contourne les limites du STG en y intégrant un nouveau type de nœud pour les caractéristiques des produits et une pénalisation des arêtes les plus anciennes. La seconde contribution est un système de recommandation basé sur l'utilisation de Link Stream Graph (LSG). Ce graphe est inspiré d'une représentation des flots de liens et a la particularité de considérer le temps de manière continue contrairement aux autres graphes de la littérature, qui soit ignore la dimension temporelle comme le graphe biparti classique (BIP), soit considère le temps de manière discontinue avec un découpage du temps en tranches comme STG.
... To calculate the similarity between the items, Karypis et al. have proposed an item-based Top-N recommendation algorithm using Cosine based similarity [36]. Jamali et al. have combined a trustbased approach to item-based CF using Pearson Correlation as the similarity metric for item similarities and weighted average method as the prediction algorithm [37]. Yang et al. [38] have presented the idea of using social networks for generating Top-N recommendation. ...
Collaborative filtering has been the most straightforward and most preferable approach in the recommender systems. This technique recommends an item to a target user from the preferences of top-k similar neighbors. In a sparse data scenario, the recommendation accuracy of the collaborative filtering degrades significantly due to the limitations of existing various similarity measures. Such constraints offer an open scope for enhancing the accuracy of optimized user-specific recommendations. Many techniques have been utilized for this, like Particle Swarm Optimization and other evolutionary collaborative filtering algorithms. The proposed approach utilizes the Apriori algorithm to form users’ profiles from the items’ ratings and categorical attributes. The user profile creation is performed using the apriori algorithm. The profile of each user involves the likes and disliking of categorical characteristics of objects by users. In the collected MovieLens dataset, the efficiency of the proposed recommendation approach is tested. The comparative results show proof that the proposed novel algorithm outperforms other prominent collaborative filtering algorithms on the MovieLens datasets based on rating prediction accuracy.
... The memory-based methods deduce ratings of a targeted user via trust propagation based on ratings of its friends [27]. For example, Jamali and Ester [28] combine TrustWalker [29] with neighborhood collaborative filtering. They first run random walks on the trust network and then perform a probabilistic item selection strategy to generate recommendations. ...
A significant remaining challenge for existing recommender systems is that users may not trust the recommender systems for either lack of explanation or inaccurate recommendation results. Thus, it becomes critical to embrace a trustworthy recommender system. This survey provides a systemic summary of three categories of trust-aware recommender systems: social-aware recommender systems that leverage users' social relationships; robust recommender systems that filter untruthful noises (e.g., spammers and fake information) or enhance attack resistance; explainable recommender systems that provide explanations of recommended items. We focus on the work based on deep learning techniques, an emerging area in the recommendation research.
... In [38], TARS (Trust Aware Recommendation System) is developed that uses the idea of dynamic trust within the users of system. A Trust Walker method is proposed in [39], which performs walk along the whole system and asks users about the ratings of their direct and indirect friends on target item. Another approach uses the concept of conditional probability for finding the level of similarity of friends on social network [40]. ...
... When the number of rated items increases, the confidence also increases and vice versa. Jamali et al. [34] propose a similarity measure, which is based on the sigmoid function. This similarity measure can weaken the similarity of small common rated items among users. ...
One of the main concerns for online shopping websites is to provide efficient and customized recommendations to a very large number of users based on their preferences. Collaborative filtering (CF) is the most famous type of recommender system method to provide personalized recommendations to users. CF generates recommendations by identifying clusters of similar users or items from the user-item rating matrix. This cluster of similar users or items is generally identified by using some similarity measurement method. Among numerous proposed similarity measure methods by researchers, the Pearson correlation coefficient (PCC) is a commonly used similarity measure method for CF-based recommender systems. The standard PCC suffers some inherent limitations and ignores user rating preference behavior (RPB). Typically, users have different RPB, where some users may give the same rating to various items without liking the items and some users may tend to give average rating albeit liking the items. Traditional similarity measure methods (including PCC) do not consider this rating pattern of users. In this article, we present a novel similarity measure method to consider user RPB while calculating similarity among users. The proposed similarity measure method state user RPB as a function of user average rating value, and variance or standard deviation. The user RPB is then combined with an improved model of standard PCC to form an improved similarity measure method for CF-based recommender systems. The proposed similarity measure is named as improved PCC weighted with RPB (IPWR). The qualitative and quantitative analysis of the IPWR similarity measure method is performed using five state-of-the-art datasets (i.e. Epinions, MovieLens-100K, MovieLens-1M, CiaoDVD, and MovieTweetings). The IPWR similarity measure method performs better than state-of-the-art similarity measure methods in terms of mean absolute error (MAE), root mean square error (RMSE), precision, recall, and F-measure.
... Zhao et al. [23] propose a social Bayesian personalized ranking model that incorporates social relationships into a pair-wise ranking model, assuming that users tend to assign higher ranks to items that their friends prefer. In [24], Jamali and Ester combine TrustWalker [4] with collaborative filtering to generate recommendations with social relationships. In [25], a collaborative ranking strategy is followed, considering how well the relevant items of users and their social friends have been ranked at the top of the list. ...
Social-based recommendation systems exploit the selections of friends to combat the data sparsity on user preferences, and improve the recommendation accuracy of the collaborative filtering strategy. The main challenge is to capture and weigh friends' preferences, as in practice they do necessarily match. In this paper, we propose a Neural Attention mechanism for Social collaborative filtering, namely NAS. We design a neural architecture, to carefully compute the non-linearity in friends' preferences by taking into account the social latent effects of friends on user behavior. In addition, we introduce a social behavioral attention mechanism to adaptively weigh the influence of friends on user preferences and consequently generate accurate recommendations. Our experiments on publicly available datasets demonstrate the effectiveness of the proposed NAS model over other state-of-the-art methods. Furthermore, we study the effect of the proposed social behavioral attention mechanism and show that it is a key factor to our model's performance.
... The advantage of our proposed Live Journal datset -RMSE approach is more trust values result in improved RMSE values. Trust network can be selected for trusted nodes based on random walk [27]. In our proposed approach, only nodes which are connected through direct or indirect trust are selected for rating prediction. ...
Deep learning is advancement in machine learning with many hidden layers. It can
provide better results even with less number of features. In deep learning, large scale data
automated learning is possible with less feature engineering. It is of high significance for
big data because of large scale of data available for training. Many applications areas such
as image processing, speech recognition and social network analysis have proven to provide
better results with the use of deep learning. Large scale of data is generated due to social
networking sites, sensor networks and business transactions. It is very difficult for user to
select any product or topic of interest, Social recommendation is active research area due to
its significance in reducing information overload. It is intelligent systems which are used to
provide suggestions to users for any product or topic. In this article, deep learning is applied
on social recommendation technique with our proposed approach based on transitive trust.
Cold start and sparsity are main limitations in social recommendation. Experiment analysis
proves that these issues are resolved by using our proposed approach incorporated with
deep learning.
... The collaborative-filtering approach recommends the items that the users with the tastes and interests similar to the target user liked in the past. The collaborative-filtering approach can be further classified into user-based [6,9,23,27,29,44,47], item-based [22,37,45], and graph-based [11,13,35,40,52,55] methods. The collaborative-filtering method suffers the cold-start problem, because the taste and interest of a new user can rarely be identified. ...
The recommender systems help users who are going through numerous items (e.g., movies or music) presented in online shops by capturing each user?s preferences on items and suggesting a set of personalized items that s/he is likely to prefer [8]. They have been extensively studied in the academic society and widely utilized in many online shops [33]. However, to the best of our knowledge, recommending items to users in price-comparison services has not been studied extensively yet, which could attract a great deal of attention from shoppers these days due to its capability to save users? time who want to purchase items with the lowest price [31]. In this paper, we examine why existing recommendation methods cannot be directly applied to price-comparison services, and propose three recommendation strategies that are tailored to price-comparison services: (1) using click-log data to identify users? preferences, (2) grouping similar items together as a user?s area of interest, and (3) exploiting the category hierarchy and keyword information of items. We implement these strategies into a unified recommendation framework based on a tripartite graph. Through our extensive experiments using real-world data obtained from Naver shopping, one of the largest price-comparison services in Korea, the proposed framework improved recommendation accuracy up to 87% in terms of precision and 129% in terms of recall, compared to the most competitive baseline.
... Trust relationships are interesting for improving recommendation, especially for cold users and cold items (users or items for which very limited information is available). Some systems incorporate trust information explicitly specified by users (Jamali and Ester, 2009;Guo et al., 2017;Pan et al., 2017), but since such explicit information is rarely available, several approaches infer implicit trust (Pitsilis and Marshall, 2004;Papagelis et al., 2005;Hwang and Chen, 2007;Lathia et al., 2008). In this section, we describe how to include these both types of trust in our framework. ...
... However, few works have focused on item ranking tasks with trust information. For example, Jamali and Ester [9] proposed the Trust-Walker method, which is possibly the first trust-based ranking method by adapting a nearest neighborhood approach to item recommendation. Zhao et al. [10] distinguished the active users' preference from the one she trusted, which is termed as SBPR. ...
Distrust based recommender systems have drawn much more attention and became widely acceptable in recent years. Previous works have investigated using trust information to establish better models for rating prediction, but there is a lack of methods using distrust relations to derive more accurate ranking-based models. In this article, we develop a novel model, named TNDBPR (Trust Neutral Distrust Bayesian Personalized Ranking), which simultaneously leverages trust, distrust, and neutral relations for item ranking. The experimental results on Epinions dataset suggest that TNDBPR by leveraging trust and distrust relations can substantially increase various performance evaluations including F1 score, AUC, Precision, Recall, and NDCG.
... Different recommendation models are compared with proposed IPG technique. TrustWalkerList finds top n recommendations by using random walk on graph of trust network [46]. Normalized cut for neighbour selection is based on collaborative filtering, and it predicts ratings with graph partitioning [47]. ...
Social big data is large scale of data due to exponential popularity of social network and social media. Researchers can use social big data and social network for their observations if they analyse those in an intelligent manner. The target of intelligent decision is to find the most credible user in social network, who has the highest influence. A very large number of users are connected in social networks, implicitly friends-of-friends or explicitly mutual friends. They are able to communicate with each other and share their likes or dislikes on different topics. If users want to analyse any topic or purchase product like movie, book, they are populated with a lot of choices. Information overload due to large number of choices available to users limits effective product selection and hence results in reduced users’ satisfaction. Recommendation models are solution for providing better suggestion to users. Product’s recommendation at Amazon, friend’s recommendation at Facebook and music recommendation at iTunes are some of the popular examples of suggestions made on the basis of user’s interests. Recommendation models ease the user by reducing search space in social network graph. The main purpose of this paper is to improve social recommendations so that better and more appropriate choices are available for users. In this paper, an efficient technique for social recommendations using hyperedge and transitive closure is proposed. Social big data is processed and analysed in the form of social graphs. User–user and user–item connections are represented in the form of matrices. We have exploited homophily so that large number of connected users have trust on each other. Our model provides better recommendation to users by leveraging increased trust. The proposed model overcomes the limitations of traditional recommender systems like sparsity, cold start. It also improves prediction accuracy. The proposed model is evaluated through different metrics like MAE, precision, recall and RMSE. Empirical analysis shows significant improvement in recommendations. We have used Mahout library for improving recommendation accuracy and also handling large volume of data. SNAP library is also used for analysis of social big graphs. The proposed recommendation model is evaluated using Epinions and FilmTrust datasets. These datasets contain user’s ratings for various products in the scale of 1–5. Through analysis it is verified that the proposed model boosts the performance significantly. We have formulated recommendation model using manipulated social graph as per our proposed technique. This manipulated graph is mentioned as influence product graph (IPG) throughout this paper. IPG increases social trust value between connected users and this effect in recommending products in an effective and efficient manner.
... This technique suggests items based on user's similar preferences. These systems can be categorized into model-based and memory-based methods [12]. The main problem in memory-based methods is calculating similarity between users and aggregation of ratings [13]. ...
... However, most previous studies on social-aware recommender systems Session 2C: Recommendation 1 CIKM'17, November 6-10, 2017, Singapore focus on rating prediction tasks, which is claimed to be less efficient in item recommendation tasks [5,33]. [14] extends the approach in [13] by combining random walks with collaborative filtering for item recommendation. The Multi-Relational Bayesian Personalized Ranking (MR-BPR) model [17], which combines multi-relational matrix factorization with the BPR framework, predicts both user feedback on items and on social relationships. ...
User feedback in the form of movie-watching history, item ratings, or product consumption is very helpful in training recommender systems. However, relatively few interactions between items and users can be observed. Instances of missing user--item entries are caused by the user not seeing the item (although the actual preference to the item could still be positive) or the user seeing the item but not liking it. Separating these two cases enables missing interactions to be modeled with finer granularity, and thus reflects user preferences more accurately. However, most previous studies on the modeling of missing instances have not fully considered the case where the user has not seen the item. Social connections are known to be helpful for modeling users' potential preferences more extensively, although a similar visibility problem exists in accurately identifying social relationships. That is, when two users are unaware of each other's existence, they have no opportunity to connect. In this paper, we propose a novel user preference model for recommender systems that considers the visibility of both items and social relationships. Furthermore, the two kinds of information are coordinated in a unified model inspired by the idea of transfer learning. Extensive experiments have been conducted on three real-world datasets in comparison with five state-of-the-art approaches. The encouraging performance of the proposed system verifies the effectiveness of social knowledge transfer and the modeling of both item and social visibilities.
... Zhao et al. [38] propose a social bayesian personalized ranking model that incorporates social relationships into a pair-wise ranking model, assuming that users tend to assign higher ranks to items that their friends prefer. In [11], Jamali and Ester combine Trust-Walker [10] with collaborative ltering to generate recommendations with social relationships. In [22], a collaborative ranking strategy is followed, considering how well the relevant items of users and their social friends have been ranked at the top of the list. ...
While users trust the selections of their social friends in recommendation systems, the preferences of friends do not necessarily match. In this study, we introduce a deep learning approach to learn both about user preferences and the social influence of friends when generating recommendations. In our model we design a deep learning architecture by stacking multiple marginalized Denoising Autoencoders. We define a joint objective function to enforce the latent representation of social relationships in the Autoencoder's hidden layer to be as close as possible to the users' latent representation when factorizing the user-item matrix. We formulate a joint objective function as a minimization problem to learn both user preferences and friends' social influence and we present an optimization algorithm to solve the joint minimization problem. Our experiments on four benchmark datasets show that the proposed approach achieves high recommendation accuracy, compared to other state-of-the-art methods.
We study a matrix completion problem which leverages a hierarchical structure of social similarity graphs as side information in the context of recommender systems. We assume that users are categorized into clusters, each of which comprises sub-clusters (or what we call “groups”). We consider a hierarchical stochastic block model that well respects practically-relevant social graphs and follows a low-rank rating matrix model. Under this setting, we characterize the information-theoretic limit on the number of observed matrix entries (i.e., optimal sample complexity) as a function of the quality of graph side information (to be detailed) by proving sharp upper and lower bounds on the sample complexity. One important consequence of this result is that leveraging the hierarchical structure of similarity graphs yields a substantial gain in sample complexity relative to the one that simply identifies different groups without resorting to the relational structure across them. Another implication of the result is when the graph information is rich, the optimal sample complexity is proportional to the number of clusters, while it nearly stays constant as the number of groups in a cluster increases. We empirically demonstrate through extensive experiments that the proposed algorithm achieves the optimal sample complexity.
Research on recommender systems has noted that the ranking of recommended items may play an important role in the performance of recommendation algorithms. To advance recommender systems research beyond the traditional approach that ranks recommended products in descending, it is crucial to understand the cognitive processes that online consumers experience when they evaluate products in a sequence. Drawing on evaluability theory and the order effects perspective, we formulate a scenario in which two products are presented sequentially and each product has two attributes, one of which can be evaluated independently while the other is difficult to evaluate without comparison. Analyses show that in two out of the three cases examined, presenting the most recommended product in the second place will result in stronger consumer purchase intentions and willingness to pay. Research hypotheses are proposed based on the results of the scenario analyses and are empirically tested through three laboratory experiments. In Study 1, evidence for the hypothesized order effects is found for the settings with randomly assigned product recommendations. In Study 2, the same effects are observed for the settings with personalized recommendations generated by a collaborative filtering algorithm. In Study 3, it is shown that such order effects also exist in terms of the recommendation strength of recommender systems. These findings provide novel insights into the behavioral implications of using recommender systems in e-commerce, shedding light on additional means of improving the design of such systems.
Bundle recommendation aims to recommend a bundle of items for a user to consume as a whole. Related work can be divided into two categories: 1) to recommend the platform's prebuilt bundles to users; 2) generate personalized bundles for users. In this work, we propose two graph neural network models, a BGCN model (short for
Bundle
Graph
Convolutional
Network
) for prebuilt bundle recommendation, and a BGGN model (short for
Bundle
Graph
Generation
Network
) for personalized bundle generation. First, BGCN unifies the user-item interaction, the user-bundle interaction and the bundle-item affiliation into a heterogeneous graph. With item nodes as the bridge, graph convolutional propagation between user and bundle nodes makes the learned representations capture the item-level semantics. Second, BGGN re-constructs bundles into graphs based on the item co-occurrence pattern and the user's supervision signal. The complex and high-order item-item relationships in the bundle graph are explicitly modeled through graph generation. Empirical results demonstrate the substantial performance gains of BGCN and BGGN, which outperforms the state-of-the-art baselines by 10.77% to 23.18% and 20.90% to 64.52%, respectively. We have released the datasets and codes at this link:
https://github.com/cjx0525/BGCN
.
As scientific and technological resources are experiencing information overload, it is quite expensive to find resources that users are interested in exactly. The personalized recommendation system is a good candidate to solve this problem, but data sparseness and the cold starting problem still prevent the application of the recommendation system. Sparse data affects the quality of the similarity measurement and consequently the quality of the recommender system. In this paper, we propose a matrix factorization recommendation algorithm based on similarity calculation(SCMF), which introduces potential similarity relationships to solve the problem of data sparseness. A penalty factor is adopted in the latent item similarity matrix calculation to capture more real relationships furthermore. We compared our approach with other 6 recommendation algorithms and conducted experiments on 5 public data sets. According to the experimental results, the recommendation precision can improve by 2% to 9% versus the traditional best algorithm. As for sparse data sets, the prediction accuracy can also improve by 0.17% to 18%. Besides, our approach was applied to patent resource exploitation provided by the wanfang patents retrieval system. Experimental results show that our method performs better than commonly used algorithms, especially under the cold starting condition.
The next-item recommendation has been in the central of interest in real-world applications such as e-commerce. However, it is challenging to infer what a user may purchase next due the complex interactions in the historical sessions and the changing semantics of an item over time. Most existing methods employ separate models to generate the general preference and the sequential patterns for the next-item recommendation without considering the interactions between the two factors or use a simple linear combination of the two factors. In this paper, we propose a deep adaptable co-embedding neural network (ACENet) to address these limitations. ACENet not only adaptably balances the combination of general preference and sequential patterns but also introduces dynamic attention for each factor in hybrid representations. Extensive experiments on two real-world datasets show the superiority of ACENet compared with other state-of-the-art methods.
Most existing studies that employ social‐trust information to solve the data sparsity issue in recommender systems assume that socially connected users have equal influence on each other. However, this assumption does not hold in practice since users and their friends may not have similar interests because social connections are multifaceted and exhibit heterogeneous strengths in different scenarios. Therefore, estimating the diverse levels of influence among entities (users/items/social connections) is very important in advancing social recommender systems. Towards this goal, we propose a new model named Social‐Trust‐Aware Variational Recommendation (SOAP‐VAE). Particularly, SOAP‐VAE leverages graph attention network techniques to capture the varying levels of influence and the complex interaction patterns among all the entities collectively and holistically. In doing so, heterogeneity among entities is obtained seamlessly. Consequently, we generate social‐trust‐aware item embedding representations in which the right level of influence has been integrated. Next, based on these rich social‐trust‐aware item representations, we formulate the first‐ever social‐trust‐aware prior in literature. Unlike priors utilized in earlier VAE‐based recommendation models, this novel prior aids in dealing with the issue of posterior‐collapse and can effectively capture the uncertainty of latent space. In effect, the model produces better latent representations, which significantly alleviates the data sparsity issue. Finally, we empirically show that SOAP‐VAE outperforms several state‐of‐the‐art baselines on three real‐world data sets.
In order to improve the recommendation accuracy and robustness of recommendation algorithm, a top-N recommendation algorithm based on reliable users is proposed in this paper. We first present a method to choose reliable users for target user using the degree of trust between users. Then based on the rating information of reliable users, we can choose candidate recommendation items using some approaches. Finally, we can provide top-N recommendation results for target user using the strategy of average aggregation. Using MovieLens dataset, we carried out a comparative experiment. The results show that the proposed algorithm improves the recommendation recall. More importantly, it has better robustness.
Social recommendation, which utilizes social relations to enhance recommender systems, has been gaining increasing attention recently with the rapid development of online social networks. Existing social recommendation methods are based on the assumption, so-called
social-trust
, that users’ preference or decision is influenced by their social-connected friends’ purchase behaviors. However, they assume that the influences of social relationships are always the same, which violates the fact that users are likely to share preference on different products with different friends. More precisely, friends’ behaviors do not necessarily affect a user’s preferences, and the influence is diverse among different items. In this paper, we contribute a new solution, CSR (short for
C
haracterized
S
ocial
R
egularization) model by designing a universal regularization term for modeling variable social influence. This regularization term captures the finely grained similarity of social-connected friends. We further introduce two variants of our model with different optimization manners. Our proposed model can be applied to both explicit and implicit interaction due to its high generality. Extensive experiments on three real-world datasets demonstrate that our CSR can outperform state-of-the-art social recommendation methods. Further experiments show that CSR can improve recommendation performance for those users with sparse social relations or behavioral interactions.
Trust aware recommender systems (TARS) are a branch of the most popular technique of recommender systems that is Collaborative Filtering. Lots of studies used trust as an enhancement factor for improve the accuracy of TARS, but the main problem of using trust is sparsity and also scalability problem for updating explicit data for new users, trust propagation used as a solution for solving the problem, but recent studies shows that using distrust values can also be useful, but same as trust and even worse, sparsity is a critical problem of dataset for using distrust. In this paper we used friend of friend and enemy of enemy concept as a way for propagating distrust and exploit new implicit relations for solving the sparsity problem and also improving the accuracy of recommendations. The results shows that propagating of distrust values is beneficial in enhancement of trust ware recommender systems.
Recommending appropriate items to users is crucial in many e-commerce platforms that contain implicit data as users' browsing, purchasing and streaming history. One common approach consists in selecting the N most relevant items to each user, for a given N, which is called top-N recommendation. To do so, recommender systems rely on various kinds of information, like item and user features, past interest of users for items, browsing history and trust between users. However, they often use only one or two such pieces of information, which limits their performance. In this paper, we design and implement GraFC2T2, a general graph-based framework to easily combine and compare various kinds of side information for top-N recommendation. It encodes content-based features, temporal and trust information into a complex graph, and uses personalized PageRank on this graph to perform recommendation. We conduct experiments on Epinions and Ciao datasets, and compare obtained performances using F1-score, Hit ratio and MAP evaluation metrics, to systems based on matrix factorization and deep learning. This shows that our framework is convenient for such explorations, and that combining different kinds of information indeed improves recommendation in general.
Traditional collaborative filtering recommendation algorithm has the problems of sparse data and limited user preference information. To deal with data sparseness problem and the unreliability phenomenon on the traditional social network recommendation. This paper presents a novel algorithm based on trust relationship reconstruction and social network delivery. This paper introduces the method of eliminating falsehood and storing truth to avoid the unreliable phenomenon and improves the accuracy of falsehood according to the user similarity formula based on the scale of contact established by users. In this paper, the problem of attack caused by the misbehaving nodes is investigated when the recommended information is disseminated in the existing trust model. In addition, a recommendation-based trust model is proposed that includes a defensive plan. This scheme employs the clustering techniques on the basis of interaction count, information Compatibility and node intimacy, in a certain period of time dynamically filter dishonest recommendation related attacks. The model has been verified in different portable and detached topologies. The network knots undergo modifications regarding their neighbors as well as frequent routes. The experimental analysis indicates correctness and robustness of the reliance system in an active MANET setting. Compared with the most advanced recommender system, the proposed recommendation algorithm in accuracy and coverage measurements show a significant improvement.
Social-aware recommender systems have been popular with the rapid growth of social media applications. Existing approaches have attempted to accommodate social information into typical Collaborative Filtering methods and achieved significant improvements. Neural networks are gaining increasing interests in information retrieval tasks. However few studies have considered applying neural network in social-aware recommendation tasks. In this paper, we aim to fill this gap and propose a social-aware neural recommender system. Extensive experiments on real-world datasets demonstrate that our model outperforms state-of-art approaches significantly.
The explosive growth of the world-wide-web and the emergence of e-commerce has led to the development of recommender systems---a personalized information filtering technology used to identify a set of items that will be of interest to a certain user. User-based collaborative filtering is the most successful technology for building recommender systems to date and is extensively used in many commercial recommender systems. Unfortunately, the computational complexity of these methods grows linearly with the number of customers, which in typical commercial applications can be several millions. To address these scalability concerns model-based recommendation techniques have been developed. These techniques analyze the user--item matrix to discover relations between the different items and use these relations to compute the list of recommendations.In this article, we present one such class of model-based recommendation algorithms that first determines the similarities between the various items and then uses them to identify the set of items to be recommended. The key steps in this class of algorithms are (i) the method used to compute the similarity between the items, and (ii) the method used to combine these similarities in order to compute the similarity between a basket of items and a candidate recommender item. Our experimental evaluation on eight real datasets shows that these item-based algorithms are up to two orders of magnitude faster than the traditional user-neighborhood based recommender systems and provide recommendations with comparable or better quality.
Recommender Systems (RS) suggest to users items they might like such as movies or songs. However they are not able to generate recommendations for users who just registered, in fact boot- strapping Recommender Systems for new users is still an open chal- lenge. While traditional RSs exploit only ratings provided by users about items, Trust-aware Recommender Systems let the user express also trust statements, i.e. their subjective opinions abou t the use- fulness of other users. In this paper we analyze the relative bene- fits of asking new users either few ratings about items or few t rust statements about other users for the purpose of bootstrapping a RS ability to generate recommendations. We run experiments on a large real world dataset derived from the online Web community of Epin- ions.com. The results clearly indicate that while traditio nal RS algo- rithms exploiting ratings on items fail for new users, askin g few trust statements to a new user is instead a very effective strategy able to quickly let the RS generate many accurate items recommendations.
Recommender Systems allow people to nd the resources they need by making use of the experiences and opinions of their nearest neigh- bours. Costly annotations by experts are replaced by a distributed pro- cess where the users take the initiative. While the collaborative approach enables the collection of a vast amount of data, a new issue arises: the quality assessment. The elicitation of trust values among users, termed \web of trust", allows a twofold enhancement of Recommender Systems. Firstly, the ltering process can be informed by the reputation of users which can be computed by propagating trust. Secondly, the trust metrics can help to solve a problem associated with the usual method of simi- larity assessment, its reduced computability. An empirical evaluation on Epinions.com dataset shows that trust propagation allows to increase the coverage of Recommender Systems while preserving the quality of pre- dictions. The greatest improuvements are achieved for new users, who provided few ratings.
Collaborative Filtering (CF) systems have been researched for over a decade as a tool to deal with information overload. At the heart of these systems are the algorithms which generate the predictions and recommendations.In this article we empirically demonstrate that two of the most acclaimed CF recommendation algorithms have flaws that result in a dramatically unacceptable user experience.In response, we introduce a new Belief Distribution Algorithm that overcomes these flaws and provides substantially richer user modeling. The Belief Distribution Algorithm retains the qualities of nearest-neighbor algorithms which have performed well in the past, yet produces predictions of belief distributions across rating values rather than a point rating value.In addition, we illustrate how the exclusive use of the mean absolute error metric has concealed these flaws for so long, and we propose the use of a modified Precision metric for more accurately evaluating the user experience.
The explosive growth of the world-wide-web and the emergence of e-commerce has led to the development of recommender systems---a personalized information filtering technology used to identify a set of N items that will be of interest to a certain user. User-based Collaborative filtering is the most successful technology for building recommender systems to date, and is extensively used in many commercial recommender systems. Unfortunately, the computational complexity of these methods grows linearly with the number of customers that in typical commercial applications can grow to be several millions. To address these scalability concerns item-based recommendation techniques have been developed that analyze the user-item matrix to identify relations between the different items, and use these relations to compute the list of recommendations. In this paper we present one such class of item-based recommendation algorithms that first determine the similarities between the various ite...
Recommender systems apply knowledge discovery techniques to the problem of making personalized recommendations for information, products or services during a live interaction. These systems, especially the k-nearest neighbor collaborative filtering based ones, are achieving widespread success on the Web. The tremendous growth in the amount of available information and the number of visitors to Web sites in recent years poses some key challenges for recommender systems. These are: producing high quality recommendations, performing many recommendations per second for millions of users and items and achieving high coverage in the face of data sparsity. In traditional collaborative filtering systems the amount of work increases with the number of participants in the system. New recommender system technologies are needed that can quickly produce high quality recommendations, even for very large-scale problems. To address these issues we have explored item-based collaborative filtering techniques. Itembased techniques first analyze the user-item matrix to identify relationships between different items, and then use these relationships to indirectly compute recommendations for users. In this paper we analyze different item-based recommendation generation algorithms. We look into different techniques for computing item-item similarities (e.g., item-item correlation vs. cosine similarities between item vectors) and different techniques for obtaining recommendations from them (e.g., weighted sum vs. regression model). Finally, we experimentally evaluate our results and compare them to the basic k-nearest neighbor approach. Our experiments suggest that item-based algorithms provide dramatically better performance than user-based algorithms, while at the same time p...
Collaborative filtering is the most popular approach to build recommender systems and has been successfully employed in many applications. However, it cannot make recommendations for so-called cold start users that have rated only a very small number of items. In addition, these methods do not know how confident they are in their recommendations. Trust-based recommendation methods assume the additional knowledge of a trust network among users and can better deal with cold start users, since users only need to be simply connected to the trust network. On the other hand, the sparsity of the user item ratings forces the trust-based approach to consider ratings of indirect neighbors that are only weakly trusted, which may decrease its precision. In order to find a good trade-off, we propose a random walk model combining the trust-based and the collaborative filtering approach for recommendation. The random walk model allows us to define and to measure the confidence of a recommendation. We performed an evaluation on the Epinions dataset and compared our model with existing trust-based and collaborative filtering methods.
Recommender Systems based on Collaborative Filtering suggest to users items they might like. However due to data sparsity of the input ratings matrix, the step of finding similar users often fails. We propose to replace this step with the use of a trust metric, an algorithm able to propagate trust over the trust network and to estimate a trust weight that can be used in place of the similarity weight. An empirical evaluation on Epinions.com dataset shows that Recommender Systems that make use of trust information are the most effective in term of accuracy while preserving a good coverage. This is especially evident on users who provided few ratings.
Collaborative Filtering recommender system, one of the most representative systems for personalized recommendations in E-commerce,
is a system assisting users in easily finding useful information. However, traditional collaborative filtering systems are
typically unable to make good quality recommendations in the situation where users have presented few opinions; this is known
as the cold start problem. In addition, the existing systems suffer some weaknesses with regard to quality evaluation: the
sparsity of the data and scalability problem. To address these issues, we present a novel approach to provide enhanced recommendation
quality supporting incremental updating of a model through the use of explicit user feedback. A model-based approach is employed
to overcome the sparsity and scalability problems. The proposed approach first identifies errors of prior predictions and
subsequently constructs a model, namely the user-item error matrix, for recommendations. An experimental evaluation on MovieLens datasets shows that the proposed method offers significant advantages both in terms of improving the recommendation quality
and in dealing with cold start users.
One of the goals in recommender systems is to recommend those items to each user that maximize the user's utility. In this study, we propose new approaches which, in conjunction with any existing recommendation technique, can improve the top-N item selection by taking into account rating variance. We empirically demonstrate how these approaches work with several recommendation techniques, increasing the accuracy of recommendations. We also show how these approaches can generate more personalized recommendations, as measured by the diversity metric. As a result, users can be given a better control to choose whether to receive recommendations with higher accuracy or higher diversity.
predicated on the belief that information filtering can be more effective when humans are involved in the filtering process. Tapestry was designed to support both content-based filtering and collaborative filtering, which entails people collaborating to help each other perform filtering by recording their reactions to documents they read. The reactions are called annotations; they can be accessed by other people’s filters. Tapestry is intended to handle any incoming stream of electronic documents and serves both as a mail filter and repository; its components are the indexer, document store, annotation store, filterer, little box, remailer, appraiser and reader/browser. Tapestry’s client/server architecture, its various components, and the Tapestry query language are described.
Viral marketing takes advantage of networks of influence among customers to inexpensively achieve large changes in behavior. Our research seeks to put it on a firmer footing by mining these networks from data, building probabilistic models of them, and using these models to choose the best viral marketing plan. Knowledge-sharing sites, where customers review products and advise each other, are a fertile source for this type of data mining. In this paper we extend our previous techniques, achieving a large reduction in computational cost, and apply them to data from a knowledge-sharing site. We optimize the amount of marketing funds spent on each customer, rather than just making a binary decision on whether to market to him. We take into account the fact that knowledge of the network is partial, and that gathering that knowledge can itself have a cost. Our results show the robustness and utility of our approach.