In real-world networks, nodes might have more than one type of relationship. Signed networks are an important class of such networks consisting of two types of relations: positive and negative. Recently, embedding signed networks has attracted increasing attention and is more challenging than classic networks since nodes are connected by paths with multi-types of links. Existing works capture the complex relationships by relying on social theories. However, this approach has major drawbacks, including the incompleteness/inaccurateness of such theories. Thus, we propose network transformation based embedding to address these shortcomings. The core idea is that rather than directly finding the similarities of two nodes from the complex paths connecting them, we can obtain their similarities through simple paths connecting their different roles. We employ this idea to build our proposed embedding technique that can be described in three steps: (1) the input directed signed network is transformed into an unsigned bipartite network with each node mapped to a set of nodes we denote as role-nodes. Each role-node captures a certain role that a node in the original network plays; (2) the network of role-nodes is embedded; and (3) the original network is encoded by aggregating the embedding vectors of role-nodes. Our experiments show the novel proposed technique substantially outperforms existing models.

Personalized hashtag recommendation for users could substantially promote user engagement in microblogging websites; users can discover microblogs aligned with their interests. However, user profiling on microblogging websites is challenging because most users tend not to generate content. Our core idea is to build a graph-based profile of users and incorporate it into hashtag recommendation. Indeed, user’s followee/follower links implicitly indicate their interests. Considering that microblogging networks are scale-free networks, to maintain the efficiency and effectiveness of the model, rather than analyzing the entire network, we model users based on their links towards hub nodes. That is, hashtags and hub nodes are projected into a shared latent space. To predict the relevance of a user to a hashtag, a projection of the user is built by aggregating the embeddings of her hub neighbors guided by an attention model and then compared with the hashtag. Classically, attention models can be trained in an end to end manner. However, due to the high complexity of our problem, we propose a novel weak supervision model for the attention component, which significantly improves the effectiveness of the model. We performed extensive experiments on two datasets collected from Twitter and Weibo, and the results confirm that our method substantially outperforms the baselines.

One of the major issues in signed networks is to use network structure to predict the missing sign of an edge. In this paper, we introduce a novel probabilistic approach for the sign prediction problem. The main characteristic of the proposed models is their ability to adapt to the sparsity level of an input network. Building a model that has an ability to adapt to the sparsity of the data has not yet been considered in the previous related works. We suggest that there exists a dilemma between local and global structures and attempt to build sparsity adaptive models by resolving this dilemma. To this end, we propose probabilistic prediction models based on local and global structures and integrate them based on the concept of smoothing. The model relies more on the global structures when the sparsity increases, whereas it gives more weights to the information obtained from local structures for low levels of the sparsity. The proposed model is assessed on three real-world signed networks, and the experiments reveal its consistent superiority over the state of the art methods. As compared to the previous methods, the proposed model not only better handles the sparsity problem, but also has lower computational complexity and can be updated using real-time data streams.

Recommender systems have been accompanied by many applications in both academia and industry. Among different algorithms used to construct a recommender system, collaborative filtering methods have attracted much attention and been used in many commercial applications. Incorporating the time into the recommendation algorithm can greatly enhance its performance. In this paper, we propose a novel time-aware model-based recommendation system. We show that future ratings of a user can be inferred from his/her rating history. We assume that there is cascade of information between the items such that rating an item can lead to other items being rated. There is indeed a hidden network structure among the items and each user tracks a sequence of items in this network. The dependencies between the items are modeled based on statistical diffusion models and the parameters are obtained through maximum-likelihood estimation. We show that under some mild assumptions, the estimation task becomes a convex optimization problem. A major advantage of the proposed method over classical recommender systems is the ability to include novel items in its recommendation lists besides providing accurate recommendations. The proposed model also results in personalized and diverse recommendations. Experimental evaluations show that the model can be trained based on the ratings of a limited number of users. Furthermore, the proposed model outperforms classical recommendation algorithms in terms of both accuracy and novelty.

Social network analysis and mining get ever-increasingly important in recent years, which is mainly due to the availability of large datasets and advances in computing systems. A class of social networks is those with positive and negative links. In such networks, a positive link indicates friendship (or trust), whereas links with a negative sign correspond to enmity (or distrust). Predicting the sign of the links in these networks is an important issue and has many applications, such as friendship recommendation and identifying malicious nodes in the network. In this manuscript, we proposed a new method for sign prediction in networks with positive and negative links. Our algorithm is based first on clustering the network into a number of clusters and then applying a collaborative filtering algorithm. The clusters are such that the number of intra-cluster negative links and inter-cluster positive links are minimal, that is, the clusters are socially balanced as much as possible (a signed graph is socially balanced if it can be divided into clusters with all positive links inside the clusters and all negative links between them). We then used similarity between the clusters (based on the links between them) in a collaborative filtering algorithm. Our experiments on a number of real datasets showed that the proposed method outperformed previous methods, including those based on social balance and status theories and one based on a machine learning framework (logistic regression in this work).

Recommender systems are in the center of network science, and they are becoming increasingly important in individual businesses for providing efficient, personalized services and products to users. Previous research in the field of recommendation systems focused on improving the precision of the system through designing more accurate recommendation lists. Recently, the community has been paying attention to diversity and novelty of recommendation lists as key characteristics of modern recommender systems. In many cases, novelty and precision do not go hand in hand, and the accuracy--novelty dilemma is one of the challenging problems in recommender systems, which needs efforts in making a trade-off between them. In this work, we propose an algorithm for providing novel and accurate recommendation to users. We consider the standard definition of accuracy and an effective self-information--based measure to assess novelty of the recommendation list. The proposed algorithm is based on item popularity, which is defined as the number of votes received in a certain time interval. Wavelet transform is used for analyzing popularity time series and forecasting their trend in future timesteps. We introduce two filtering algorithms based on the information extracted from analyzing popularity time series of the items. The popularity-based filtering algorithm gives a higher chance to items that are predicted to be popular in future timesteps. The other algorithm, denoted as a novelty and population-based filtering algorithm, is to move toward items with low popularity in past timesteps that are predicted to become popular in the future. The introduced filters can be applied as adds-on to any recommendation algorithm. In this article, we use the proposed algorithms to improve the performance of classic recommenders, including item-based collaborative filtering and Markov-based recommender systems. The experiments show that the algorithms could significantly improve both the accuracy and effective novelty of the classic recommenders.