Ming-Sheng Shang

Empirical analysis of web-based user-object bipartite networks

Understanding the structure and evolution of web-based user-object networks is a significant task since they play a crucial role in e-commerce nowadays. This letter reports the empirical analysis on two large-scale web sites, audioscrobbler.com and del.icio.us, where users are connected with music groups and bookmarks, respectively. The degree distributions and degree-degree correlations for both users and objects are reported. We propose a new index, named collaborative similarity, to quantify the diversity of tastes based on the collaborative selection. Accordingly, the correlation between degree and selection diversity is investigated. We report some novel phenomena well characterizing the selection mechanism of web users and outline the relevance of these phenomena to the information recommendation problem.

Collaborative filtering with diffusion-based similarity on tripartite graphs

Collaborative tags are playing a more and more important role for the organization of information systems. In this paper, we study a personalized recommendation model making use of the ternary relations among users, objects and tags. We propose a measure of user similarity based on his preference and tagging information. Two kinds of similarities between users are calculated by using a diffusion-based process, which are then integrated for recommendation. We test the proposed method in a standard collaborative filtering framework with three metrics: ranking score, Recall and Precision, and demonstrate that it performs better than the commonly used cosine similarity.

The reinforcing influence of recommendations on global diversification

Recommender systems are promising ways to filter the abundant information in modern society. Their algorithms help individuals to explore decent items, but it is unclear how they distribute popularity among items. In this paper, we simulate successive recommendations and measure their influence on the dispersion of item popularity by Gini coefficient. Our result indicates that local diffusion and collaborative filtering reinforce the popularity of hot items, widening the popularity dispersion. On the other hand, the heat conduction algorithm increases the popularity of the niche items and generates smaller dispersion of item popularity. Simulations are compared to mean-field predictions. Our results suggest that recommender systems have reinforcing influence on global diversification. Finally, the study of the hybrid method of mass diffusion and heat conduction reveals that the influence of recommender systems is actually controllable.

Relevance is more significant than correlation: Information filtering on sparse data

In some recommender systems where users can vote objects by ratings, the similarity between users can be quantified by a benchmark index, namely the Pearson correlation coefficient, which reflects the rating correlations. Another alternative way is to calculate the similarity based solely on the relevance information, namely whether a user has voted an object. The former one uses more information than the latter, and is intuitively expected to give more accurate rating predictions under the standard collaborative filtering framework. However, according to the extensive experimental analysis, this letter reports the opposite results that the latter method, making use of only the relevance information, can outperform the former method, especially when the data set is sparse. Our finding challenges the routine knowledge on information filtering, and suggests some alternatives to address the sparsity problem.

Extracting the information backbone in online system

Information overload is a serious problem in modern society and many solutions such as recommender system have been proposed to filter out irrelevant information. In the literature, researchers have been mainly dedicated to improving the recommendation performance (accuracy and diversity) of the algorithms while they have overlooked the influence of topology of the online user-object bipartite networks. In this paper, we find that some information provided by the bipartite networks is not only redundant but also misleading. With such “less can be more” feature, we design some algorithms to improve the recommendation performance by eliminating some links from the original networks. Moreover, we propose a hybrid method combining the time-aware and topology-aware link removal algorithms to extract the backbone which contains the essential information for the recommender systems. From the practical point of view, our method can improve the performance and reduce the computational time of the recommendation system, thus improving both of their effectiveness and efficiency.

CAN DISSIMILAR USERS CONTRIBUTE TO ACCURACY AND DIVERSITY OF PERSONALIZED RECOMMENDATION

Recommender systems are becoming a popular and important set of personalization techniques that assist individual users with navigating through the rapidly growing amount of information. A good recommender system should be able to not only find out the objects preferred by users, but also help users in discovering their personalized tastes. The former corresponds to high accuracy of the recommendation, while the latter to high diversity. A big challenge is to design an algorithm that provides both highly accurate and diverse recommendation. Traditional recommendation algorithms only take into account the contributions of similar users, thus, they tend to recommend popular items for users ignoring the diversity of recommendations. In this paper, we propose a recommendation algorithm by considering both the effects of similar and dissimilar users under the framework of collaborative filtering. Extensive analyses on three datasets, namely MovieLens, Netflix and Amazon, show that our method performs much better than the standard collaborative filtering algorithm for both accuracy and diversity.

Iterative resource allocation based on propagation feature of node for identifying the influential nodes

Abstract The identification of the influential nodes in networks is one of the most promising domains. In this paper, we present an improved iterative resource allocation (IIRA) method by considering the centrality information of neighbors and the influence of spreading rate for a target node. Comparing with the results of the Susceptible Infected Recovered (SIR) model for four real networks, the IIRA method could identify influential nodes more accurately than the tradition IRA method. Specially, in the Erdos network, Kendalls tau could be enhanced 23% when the spreading rate is 0.12. In the Protein network, Kendalls tau could be enhanced 24% when the spreading rate is 0.08.

SIMILARITY-BASED CLASSIFICATION IN PARTIALLY LABELED NETWORKS

Two main difficulties in the problem of classification in partially labeled networks are the sparsity of the known labeled nodes and inconsistency of label information. To address these two difficulties, we propose a similarity-based method, where the basic assumption is that two nodes are more likely to be categorized into the same class if they are more similar. In this paper, we introduce ten similarity indices defined based on the network structure. Empirical results on the co-purchase network of political books show that the similarity-based method can, to some extent, overcome these two difficulties and give higher accurate classification than the relational neighbors method, especially when the labeled nodes are sparse. Furthermore, we find that when the information of known labeled nodes is sufficient, the indices considering only local information can perform as good as those global indices while having much lower computational complexity.

Collaborative filtering based on multi-channel diffusion

In this paper, by applying a diffusion process, we propose a new index to quantify the similarity between two users in a user–object bipartite graph. To deal with the discrete ratings on objects, we use a multi-channel representation where each object is mapped to several channels with the number of channels being equal to the number of different ratings. Each channel represents a certain rating and a user having voted an object will be connected to the channel corresponding to the rating. Diffusion process taking place on such a user–channel bipartite graph gives a new similarity measure of user pairs, which is further demonstrated to be more accurate than the classical Pearson correlation coefficient under the standard collaborative filtering framework.

A Novel Approach to Extracting Non-Negative Latent Factors From Non-Negative Big Sparse Matrices

An inherently non-negative latent factor model is proposed to extract non-negative latent factors from non-negative big sparse matrices efficiently and effectively. A single-element-dependent sigmoid function connects output latent factors with decision variables, such that non-negativity constraints on the output latent factors are always fulfilled and thus successfully separated from the training process with respect to the decision variables. Consequently, the proposed model can be easily and fast built with excellent prediction accuracy. Experimental results on an industrial size sparse matrix are given to verify its outstanding performance and suitability for industrial applications.