User-item Rating Matrix Research Articles

A hybrid recommendation system based on profile expansion technique to alleviate cold start problem

Recommender systems are one of the information filtering tools which can be employed to find interest items of users. Collaborative filtering is one of the recommendation methods to provide suggestions for target users based on the ratings of like-interest users. This method suffers from some shortcomings such as cold start problem leading to reduce the performance of recommender system in predicting unseen items. In this paper, we propose a hybrid recommendation method based on profile expansion technique to alleviate cold start problem in recommender systems. For this purpose, we take into consideration user’s demographic data (e.g. age, gender, and occupation) beside user’s rating data in order to enrich the neighborhood set of users. Specifically, two different strategies are used to enrich the rating profile of users by adding some additional ratings to them. The proposed rating profile expansion mechanism has a significant effect on the performance improvement of recommender systems especially when they are facing with cold start problem. The reason behind this claim is that the proposed mechanism makes a denser user-item rating matrix than the original one by adding some additional ratings to it. Obviously, providing a rating profile with further ratings for the target user leads to alleviate cold start problem in recommender systems. The expanded rating profiles are used to calculate similarity values between users and predict unseen items. The results of experiments demonstrate that the proposed method can achieve better performance than the other recommendation methods in terms of accuracy and rate coverage measures.

A new user similarity measure in a new prediction model for collaborative filtering

The Recommender Systems (RSs) based on the performance of Collaborative filtering (CF) depends on similarities among users or items obtained by a user-item rating matrix. The conventional measures such as the Pearson correlation coefficient (PCC), cosine (COS), and Jaccard (JACC) provide a varied and dissimilar value when the ratings between the users lie in the positive and negative side of the rating scale. These measures are also not very effective when there is sparsity in the rating matrix of the user-item. These problems are addressed by the Proximity-Impact-Popularity (PIP) similarity measure. Even though the PIP method provides an improved solution for this problem, the range of values for each component in PIP is very high. To address this issue and to improve the performance of a CF-based RS, a modified proximity-impact-popularity (MPIP) similarity measure is introduced. The expression is designed to get PIP values within the range of 0 to 1. A modified prediction expression is proposed to predict the available and unavailable ratings by combining user- and item-related components. The proposed method is tested by using various benchmark datasets. The size of the user-item sparse matrix varies to compare the performance of the methods in terms of mean absolute error, root mean squared error, precision, recall, and F1-measure. The performance of the proposed method is statistically tested through the Friedman and McNemer test. The results obtained by using the evaluation criteria indicate that the proposed method provides a better solution than the conventional methods. The statistical analysis reveals that the proposed method provides minimum MAE and RMSE values. Similarly, it also provides a maximum F1-measure for all the sub-problems.

Local ranking and global fusion for personalized recommendation

Most of top-n personalized recommendation algorithms perform a kind of global learning-to-rank based on the original user–item rating matrix, yet with an implicit presumption of a unified feature space for all users (all items). However, we argue that such a presumption is a limiting factor for better recommendation, because features learned in a global space could be too coarse-grained. As such, ranking a recommendation list against a unified space might be dominated by only a few significant aspects of users’ interests.In this paper, we propose a novel local ranking and global fusion (LRGF) framework for personalized recommendation, which allows using multiple fine-grained feature spaces for learning-to-rank. In LRGF, we first adopt two-stage random walk on the established user graph and item graph for selecting those mostly related users and items, so as to form multiple local user–item groups. Next, we propose a pairwise gap-aware Bayesian personalized ranking to obtain local ranking lists for each local user–item group. Finally, a global decision fusion is proposed to obtain the final top-n recommendation list for each user only from her local ranking lists. Experiments on the three real-world datasets show that the proposed LRGF outperforms the state-of-the-art schemes in terms of higher mean average precision and normalized discounted cumulative gain.

Unsupervised and supervised methods for the detection of hurriedly created profiles in recommender systems

Recommender systems try to provide users with accurate personalized suggestions for items based on an analysis of previous user decisions and the decisions made by other users. These systems suffer from profile injection attacks, where malicious profiles are generated in order to promote or demote a particular item introducing abnormal ratings. The problem of automatic detection of such malicious profiles has been recently addressed by a great number of authors in the literature using supervised and unsupervised approaches. In this paper, we propose a framework to identify anomalous rating profiles, where each attacker (outlier) hurriedly creates profiles that inject into the system an unspecified combination of random ratings and specific ratings, without any prior knowledge of the existing ratings. This attack is a superset of the two different attacks (Uniform and Delta) proposed in Harper et al. (ACM Trans Interact Intell Syst 5(4):19, 2016) making the attack model more realistic and its detection more challenging. The proposed detection method is based on several attributes related to the unpredictable behavior of the outliers in a validation set, on the user-item rating matrix, on the similarity between users and on the filler items. In this work, we propose a new attribute (RIS) to capture the randomness in item selection of the abnormal profiles. In this work, three different systems are proposed: (1) a probabilistic framework that estimates the probability of a user to be an outlier by combining several features in a completely unsupervised way. (2) An unsupervised clustering system based on the k-means algorithm that automatically spots the spurious profiles. (3) A supervised framework that uses a random forest classifier for cases where labeling sample data is available. Experimental results on the MovieLens and the Small Netflix datasets demonstrate the high performance of the proposed methods as well as the discrimination accuracy of the proposed features.

A Deep Graph Neural Network-Based Mechanism for Social Recommendations

Nowadays, the issue of information overload is gradually gaining exposure in the Internet of Things (IoT), calling for more research on recommender system in advance for industrial IoT scenarios. With the ever-increasing prevalence of various social networks, social recommendations (SoR) will certainly become an integral application that provides more feasibly personalized information service for future IoT users. However, almost all of the existing research managed to explore and quantify correlations between user preferences and social relationships, while neglecting the correlations among item features which could further influence the topologies of some social groups. To tackle with this challenge, in this article, a deep graph neural network-based social recommendation framework (GNN-SoR) is proposed for future IoTs. First, user and item feature spaces are abstracted as two graph networks and respectively encoded via the graph neural network method. Next, two encoded spaces are embedded into two latent factors of matrix factorization to complete missing rating values in a user-item rating matrix. Finally, a large amount of experiments are conducted on three real-world data sets to verify the efficiency and stability of the proposed GNN-SoR.

Leveraging User Comments for Recommendation in E-Commerce

Collaborative filtering recommender systems traditionally recommend products to users solely based on the given user-item rating matrix. Two main issues, data sparsity and scalability, have long been concerns. In our previous work, an approach was proposed to address the scalability issue by clustering the products using the content of the user-item rating matrix. However, it still suffers from these concerns. In this paper, we improve the approach by employing user comments to address the issues of data sparsity and scalability. Word2Vec is applied to produce item vectors, one item vector for each product, from the comments made by users on their previously bought goods. Through the user-item rating matrix, the user vectors of all the customers are produced. By clustering, products and users are partitioned into item groups and user groups, respectively. Based on these groups, recommendations to a user can be made. Experimental results show that both the inaccuracy caused by a sparse user-item rating matrix and the inefficiency due to an enormous amount of data can be much alleviated.

A Two-Stage Rating Prediction Approach Based on Matrix Clustering on Implicit Information

Traditional matrix factorization (MF) methods take a global view on the user-item rating matrix to conduct matrix decomposition for rating approximation. However, there is an inherent structure in the user-item rating matrix and a local correspondence between user clusters and item clusters as the users induce the items and the items imply the users in a recommendation system. This article proposes a novel approach called two-stage rating prediction (TS-RP) to matrix clustering with implicit information. In the first stage, implicit feedback is used to discover the inherent structure of the user-item rating matrix by spectral clustering. In the second stage, we conduct rating prediction on the dense blocks of explicit information of user-item clusters discovered in the first stage. The proposed TS-RP approach can not only alleviate the data sparsity problem in recommendation but also increase the computation scalability. Experiments on the MovieLens-100K data set demonstrate that the proposed TS-RP approach performs better than most state-of-the-art methods of rating prediction based on MF in terms of recommendation accuracy and computation complexity.

Trust information network in social Internet of things using trust-aware recommender systems

Social Internet of things is one of the most up-to-date research issues in the applications of Internet of things technologies. In social Internet of things, accuracy and reliability are standard features to discerning decisions. We assume that decision support systems based on social Internet of things could leverage research from recommender systems to achieve more stable performance. Therefore, we propose a trust-aware recommender systems suitable for social Internet of things. Trust-aware recommender systems adapt the concept of social networking service and utilize social interaction information. Trust information not only improves recommender systems from opinion spam problems but also more accurately predicts users’ preferences. We confirm that the performance of a recommender system becomes more improved when implicit trust is able to satisfy the properties of trust in the social Internet of things environment. The structure and amount of social link information are context-sensitive, so applying the concept of trust into social Internet of things environments requires a method to optimize implicit and explicit trust with minimal social link information. Our proposed method configures an asymmetric implicit trust network utilizing user–item rating matrix and transforms trust propagation metrics for a directional and weighted trust network. Through experiments, we confirm that the proposed methods enable higher accuracy and wider coverage compared to the existing recommendation methods.

Recommending Curated Content Using Implicit Feedback

Matrix factorization (MF) which is a Collaborative filtering (CF) based model, is widely used in the recommendation systems (RS). For our experiment, we collected data from a company's internal web site where curated contents are published and pushed to the employees. However, the size of the dataset is small and interaction data is also limited. We got a sparse matrix when we generated a user-item rating matrix. We have used Multi-Layer Perceptron (MLP) to calculate the rating scores from the implicit feedbacks. However, on this sparse dataset traditional content only or CF-only RSs do not work well. Here, we propose ahybrid RS that incorporates content similarity scores into an MLP-based MF-model. To integrate the content similarity scores into the MF, we have defined an objective function based on a regularization term. The experimental result shows that our proposed model demonstrates a better result than the traditional MF-based models.

A Hybrid Collaborative Filtering Recommendation Algorithm Based on User Attributes and Matrix Completion

Collaborative filtering is a popular strategy in recommendation system. Traditional collaborative filtering relies on the user-item rating matrix that encodes the individual ratings of users for items to make recommendations. However, in the real-world, the rating matrix is highly sparse, and many new users do not have rating records, thus traditional collaborative filtering could not provide satisfactory recommendations. To alleviate this issue, we propose a hybrid algorithm that utilizes LMaFit to complete rating matrix, reducing the degree of sparsity, and provides a hybrid user-similarity to supply a good support for recommending to new users in the condition of cold start. Extensive experiment results on real-world datasets show the proposed algorithm has a better performance than other methods.

An enhanced social matrix factorization model for recommendation based on social networks using social interaction factors

Recommender systems are recently becoming more significant in the age of rapid development of Internet technology and pervasive computing due to their ability in making appropriate choices to users. Collaborative filtering is one of the most successful recommendation techniques, which recommends items to an active user based on past ratings from like-minded users. However, the user-item rating matrix, namely one of the inputs to the recommendation algorithm, is often highly sparse, thus collaborative filtering may lead to the poor recommendation. To solve this problem, social networks can be employed to improve the accuracy of recommendations. Some of the social factors have been used in recommender system, but have not been fully considered. In this paper, we fuse personal cognition behavior, cognition relationships between users, and time decay factor for rated items into a unified probabilistic matrix factorization model and propose an enhanced social matrix factorization approach for personalized recommendation using social interaction factors. In this study, we integrate propagation enhancement, common user relationship enhancement, and common interest enhancement into social relationship between users, and propose a novel trust relationship calculation to alleviate the negative impact of sparsity of data rating. The proposed model is compared with the existing social recommendation algorithms on real world datasets including the Epinions and Movielens datasets. Experimental results demonstrate that our proposed approach achieves superior performance to the other recommendation algorithms.

Learning to Make Document Context-Aware Recommendation with Joint Convolutional Matrix Factorization

Context-aware recommendation (CR) is the task of recommending relevant items by exploring the context information in online systems to alleviate the data sparsity issue of the user-item data. Prior methods mainly studied CR by document-based modeling approaches, that is, making recommendations by additionally utilizing textual data such as reviews, abstracts, or synopses. However, due to the inherent limitation of the bag-of-words model, they cannot effectively utilize contextual information of the documents, which results in a shallow understanding of the documents. Recent works argued that the understanding of document context can be improved by the convolutional neural network (CNN) and proposed the convolutional matrix factorization (ConvMF) to leverage the contextual information of documents to enhance the rating prediction accuracy. However, ConvMF only models the document content context from an item view and assumes users are independent and identically distributed (i.i.d). But in reality, as we often turn to our friends for recommendations, the social relationship and social reviews are two important factors that can change our mind most. Moreover, users are more inclined to interact (buy or click) with the items that they have bought (or clicked). The relationships among items are also important factors that can impact the user’s final decision. Based on the above observations, in this work, we target CR and propose a joint convolutional matrix factorization (JCMF) method to tackle the encountered challenges, which jointly considers the item’s reviews, item’s relationships, user’s social influence, and user’s reviews in a unified framework. More specifically, to explore items’ relationships, we introduce a predefined item relation network into ConvMF by a shared item latent factor and propose a method called convolutional matrix factorization with item relations (CMF-I). To consider user’s social influence, we further integrate the user’s social network into CMF-I by sharing the user latent factor between user’s social network and user-item rating matrix, which can be treated as a regularization term to constrain the recommendation process. Finally, to model the document contextual information of user’s reviews, we exploit another CNN to learn user’s content representations and achieve our final model JCMF. We conduct extensive experiments on the real-world dataset from Yelp. The experimental results demonstrate the superiority of JCMF compared to several state-of-the-art methods in terms of root mean squared error (RMSE) and mean average error (MAE).

Research on the filtering recommendation technology of network information based on big data environment

Aiming at the problems of long recommendation time and low accuracy in traditional filtering recommendation methods for network information, a filtering recommendation method based on trusted user rating of network information is proposed. This paper analyses the relationship between the evaluation time of network users and the change of users' interest in the item in different time periods, builds the time forgetting model and time window model of network information, generates time filtering function, filters the historical score of the item by time filtering function, obtains the similarity of users' score, and constructs the trusted user-item rating matrix of network information, to generate the rating recommendation list of target user for network information prediction, and then complete recommendation. The experimental results show that the proposed method has shorter recommendation completion time and higher accuracy.

Research on the filtering recommendation technology of network information based on big data environment

Aiming at the problems of long recommendation time and low accuracy in traditional filtering recommendation methods for network information, a filtering recommendation method based on trusted user rating of network information is proposed. This paper analyses the relationship between the evaluation time of network users and the change of users' interest in the item in different time periods, builds the time forgetting model and time window model of network information, generates time filtering function, filters the historical score of the item by time filtering function, obtains the similarity of users' score, and constructs the trusted user-item rating matrix of network information, to generate the rating recommendation list of target user for network information prediction, and then complete recommendation. The experimental results show that the proposed method has shorter recommendation completion time and higher accuracy.

An Improved Product Recommendation Method for Collaborative Filtering

Collaborative filtering (CF) is the most commonly used technique for online recommendations. CF works by computing the interests of a user by gathering preferences or taste information of other users. In this technique, similar users or items are discovered by exploring the user-item rating matrix. Based on the computed similarity, a prediction is made for the unknown or new product. There are many similarity computation methods, such as the Pearson correlation coefficient (PCC), Jaccard, Mean square difference, Cosine, etc. however, the accuracy of product recommendations using these methods is not very promising. This work introduces an improved product recommendation method for collaborative filtering, which is based on the triangle similarity. However, the downside of triangle similarity is that it only considers the common ratings of users. The proposed similarity measure not only focuses on common ratings but also consider the ratings of those items that are not commonly rated from pairs of users. The obtained similarity is further complemented with the user rating preference (URP) behavior in giving rating preferences. To evaluate the accuracy, experiments are performed on the six commonly used datasets in the field of CF. Experimental results prove that the proposed similarity measure performs well as compared to the existing similarity measures.

Popularised Similarity Function for Effective Collaborative Filtering Recommendations

The existing similarity functions use the user-item rating matrix to process similar neighbours that can be used to predict ratings to the users. However, the functions highly penalise high popular items which lead to predicting items that may not be of interest to active users due to the punishment function employed. The functions also reduce the chances of selecting less popular items as similar neighbours due to the items with common ratings used. In this article, a popularised similarity function (pop_sim) is proposed to provide effective recommendations to users. The pop_sim function introduces a modified punishment function to minimise the penalty on high popular items. The function also employs a popularity constraint which uses ratings threshold to increase the chances of selecting less popular items as similar neighbours. The experimental studies indicate that the proposed pop_sim is effective in improving the accuracy of the rating prediction in terms of not only lowering the MAE but also the RMSE.

Local matrix approximation via automatic anchor selection and asymmetric neighbor inclusion based on feature divergence measure

Due to the ever-increasing users and items, the global matrix factorization (MF) over a large yet sparse user-item rating matrix no longer suffices for accurate rating approximations. This paper studies the problem of how to obtain many local matrices for improving the global rating approximation via averaging local MF results. In our scheme, we propose to use a feature divergence measure to compute the asymmetric distances in between users and items. To obtain local matrix approximation, we first propose to use a density-based clustering algorithm to automatically select a number of user-item pairs as anchors for self-adapting different datasets. Next for each anchor, we recruit its neighboring users and items with small feature divergences to the anchor, thus forming each local matrix. A kernel-based approach is then applied for local matrix approximation and a weighted averaging is used to obtain the global matrix rating approximations. Our experiments on the three real-world datasets show that the proposed solution can outperform the state-of-the-art schemes in terms of lower approximation errors and higher coverage ratios.

Unifying user similarity and social trust to generate powerful recommendations for smart cities using collaborating filtering-based recommender systems

Recommender systems can improve the quality of life in smart cities by presenting personalized services to the community. Such systems maintain a database of user profiles for producing recommendations for a specific user. The collaborative filtering (CF) approach used in these systems has become a benchmark approach for generating recommendations for interested users because it can provide “out of the box” solutions. These CF-based approaches first construct a user–item rating matrix and then exploit similarity methods. These approaches suffer from scalability, sparsity, and cold user conditions, which consequently result in the poor recommendation accuracy of these systems. To enhance the accuracy of recommender systems, social trust can play a vital role because people tend to interact with a system or respond positively to recommendations that originate from their social trustworthy friends. The proposed unified approach of this article uses explicit trust, implicit trust, and user preference similarity to create a unified rating profile for the target user to produce more powerful and accurate recommendations. The proposed unified approach also enhances the recommendation performance of CF-based recommender systems when only a limited set of ratings is available. Experiments are performed on three publicly available datasets which are FilmTrust, CiaoDVD, and Epinions. Comparison of obtained results is made with traditional similarity measures as well as up-to-date trust-based approaches. The results show that the proposed unified approach is superior to existing approaches in terms of both predictive and classification-based accuracy measures.

Destructure-and-restructure matrix approximation

A matrix approximation is used to predict the missing values of a matrix, which has many applications in recommender systems. After predicting the missing values of a user-item rating matrix, we can recommend unrated items to a user according to their approximated values. This paper proposes a new approximation scheme for large and sparse global matrices, which is called a destructure-and-restructure matrix approximation. The basic idea is to first destructure a global matrix into many local matrices, and then restructure it from local matrix approximations. However, distance computation is a challenging issue in matrix destructure because no prior knowledge about the most appropriate feature vectors and distance measures are available. To deal with this issue, we propose a novel scheme that does not require any distance computation for local matrix construction, which is based on the application of convergence probabilities of a graph random walk. At first, a user-item bipartite graph is built from the global matrix. After performing random walk on the bipartite graph, we select several user-item pairs as anchors. Then, another random walk with restart is applied to construct local matrices from anchors. For each local matrix, we propose a weighted matrix factorization that is based on the rating distribution of the training data. In the restructure process, we first compute the approximation credibility of each local matrix, and we then finally obtain the global matrix approximation weighted from credible local approximations. Our experiments on five real-world datasets show that the proposed solution outperforms the state-of-the-art schemes in terms of lower prediction errors and higher coverage ratios.

Adaptive Local Low-rank Matrix Approximation for Recommendation

Low-rank matrix approximation (LRMA) has attracted more and more attention in the community of recommendation. Even though LRMA-based recommendation methods (including Global LRMA and Local LRMA) obtain promising results, they suffer from the complicated structure of the large-scale and sparse rating matrix, especially when the underlying system includes a large set of items with various types and a huge amount of users with diverse interests. Thus, they have to predefine the important parameters, such as the rank of the rating matrix and the number of submatrices. Moreover, most existing Local LRMA methods are usually designed in a two-phase separated framework and do not consider the missing mechanisms of rating matrix. In this article, a non-parametric unified Bayesian graphical model is proposed for A daptive Lo cal low-rank M atrix A pproximation ( ALoMA ). ALoMA has ability to simultaneously identify rating submatrices, determine the optimal rank for each submatrix, and learn the submatrix-specific user/item latent factors. Meanwhile, the missing mechanism is adopted to characterize the whole rating matrix. These four parts are seamlessly integrated and enhance each other in a unified framework. Specifically, the user-item rating matrix is adaptively divided into proper number of submatrices in ALoMA by exploiting the Chinese Restaurant Process. For each submatrix, by considering both global/local structure information and missing mechanisms, the latent user/item factors are identified in an optimal latent space by adopting automatic relevance determination technique. We theoretically analyze the model’s generalization error bounds and give an approximation guarantee. Furthermore, an efficient Gibbs sampling-based algorithm is designed to infer the proposed model. A series of experiments have been conducted on six real-world datasets ( Epinions , Douban , Dianping , Yelp , Movielens (10M), and Netflix ). The results demonstrate that ALoMA outperforms the state-of-the-art LRMA-based methods and can easily provide interpretable recommendation results.