Xuzhen Zhu

Personalized recommendation with corrected similarity

Personalized recommendation has attracted a surge of interdisciplinary research. Especially, similarity-based methods in applications of real recommendation systems have achieved great success. However, the computations of similarities are overestimated or underestimated, in particular because of the defective strategy of unidirectional similarity estimation. In this paper, we solve this drawback by leveraging mutual correction of forward and backward similarity estimations, and propose a new personalized recommendation index, i.e., corrected similarity based inference (CSI). Through extensive experiments on four benchmark datasets, the results show a greater improvement of CSI in comparison with these mainstream baselines. And a detailed analysis is presented to unveil and understand the origin of such difference between CSI and mainstream indices.

Personalized recommendation based on unbiased consistence

Recently, in physical dynamics, mass-diffusion–based recommendation algorithms on bipartite network provide an efficient solution by automatically pushing possible relevant items to users according to their past preferences. However, traditional mass-diffusion–based algorithms just focus on unidirectional mass diffusion from objects having been collected to those which should be recommended, resulting in a biased causal similarity estimation and not-so-good performance. In this letter, we argue that in many cases, a users interests are stable, and thus bidirectional mass diffusion abilities, no matter originated from objects having been collected or from those which should be recommended, should be consistently powerful, showing unbiased consistence. We further propose a consistence-based mass diffusion algorithm via bidirectional diffusion against biased causality, outperforming the state-of-the-art recommendation algorithms in disparate real data sets, including Netflix, MovieLens, Amazon and Rate Your Music.

Information filtering based on corrected redundancy-eliminating mass diffusion

Methods used in information filtering and recommendation often rely on quantifying the similarity between objects or users. The used similarity metrics often suffer from similarity redundancies arising from correlations between objects’ attributes. Based on an unweighted undirected object-user bipartite network, we propose a Corrected Redundancy-Eliminating similarity index (CRE) which is based on a spreading process on the network. Extensive experiments on three benchmark data sets—Movilens, Netflix and Amazon—show that when used in recommendation, the CRE yields significant improvements in terms of recommendation accuracy and diversity. A detailed analysis is presented to unveil the origins of the observed differences between the CRE and mainstream similarity indices.

Link prediction based on nonequilibrium cooperation effect

Link prediction in complex networks has become a common focus of many researchers. But most existing methods concentrate on neighbors, and rarely consider degree heterogeneity of two endpoints. Node degree represents the importance or status of endpoints. We describe the large-degree heterogeneity as the nonequilibrium between nodes. This nonequilibrium facilitates a stable cooperation between endpoints, so that two endpoints with large-degree heterogeneity tend to connect stably. We name such a phenomenon as the nonequilibrium cooperation effect. Therefore, this paper proposes a link prediction method based on the nonequilibrium cooperation effect to improve accuracy. Theoretical analysis will be processed in advance, and at the end, experiments will be performed in 12 real-world networks to compare the mainstream methods with our indices in the network through numerical analysis.

Hybrid influence of degree and H-index in the link prediction of complex networks

Previous link prediction researchers paid more attention to the delivery ability of paths between two unlinked endpoints, but less to the influences of endpoints. In this letter, we uncover that synthesizing degree and H -index as the hybrid influences of endpoints can more reliably capture such endpoints with great and extensive maximum connected subgraph, which can more possibly attract other unlinked endpoints. In addition, the influence involving small heterogeneity of degree and H -index can further improve the accuracy of link prediction. Based on the hybrid influences of endpoints, we propose link prediction methods to explore the mechanism of link evolution. Extensive experiments on twelve real datasets suggest that the proposed methods can remarkably promote accuracy of link prediction.

Personalized Recommendation via Suppressing Excessive Diffusion

Efficient recommendation algorithms are fundamental to solve the problem of information overload in modern society. In physical dynamics, mass diffusion is a powerful tool to alleviate the long-standing problems of recommendation systems. However, popularity bias and redundant similarity have not been adequately studied in the literature, which are essentially caused by excessive diffusion and will lead to similarity estimation deviation and recommendation performance degradation. In this paper, we penalize the popular objects by appropriately dividing the popularity of objects and then leverage the second-order similarity to suppress excessive diffusion. Evaluation on three real benchmark datasets (MovieLens, Amazon, and RYM) by 10-fold cross-validation demonstrates that our method outperforms the mainstream baselines in accuracy, diversity, and novelty.