Showing papers by "Xiaoheng Deng published in 2015"

Credit Distribution and influence maximization in online social networks using node features

Abstract: Influence maximization is a problem of identifying a small set of highly influential individuals such that obtaining the maximized influence spread after propagation in social networks. How to evaluate the influence is essential to solve the influence maximization problem. Meanwhile, finding out influence propagation paths is one of key factors in the assessment of influence spread. However, since most of existent models and algorithms use degrees to simplify the activation probability on edges, node features are always ignored in the evaluation of influence ability for different users. In this paper, besides the node features, the Credit Distribution (CD) model is extended to incorporate the time-critical aspect of influence in social networks. After assigning credit along with the action propagation paths, we pick up the nodes which have maximal marginal gain in each iteration to form the seed set. The experiments we performed on real online social networks demonstrate that our approach is efficiency and reasonability for identifying seed sets, and the influence spread prediction by our approach is more accurately than that of original algorithm which disregards the node features in the influence evaluation and diffusion.

Credit Distribution for Influence Maximization in Online Social Networks with Time Constraint

Abstract: Considering the time constraint, influence maximization with time constraint (IMTC) is a problem of identifying several maximum influential individuals as seed nodes who will influence others and lead to the largest number of adoption in an expected sense. Associated with probabilities of events and the radio of information gain, we propose an optimized approach to evaluate the activation probability synthetically. As the credit which indicates the strength of influence given to adjacent neighbors is depended on the optimized activation probability (OAP), we also extend the Credit Distribution (CD) model by restricting the scope of credit distribution with the time-delay aspect of influence diffusion in online social networks. Furthermore, the time obstacle caused by repeated attempts is converted to length of the action propagation augmented paths (APAP). The simulations and experiments implemented on real datasets manifest that our approach is more effectively and efficiently in identifying seed nodes and predicting influence diffusion compared with other related approaches.

Finding overlapping communities with random walks on line graph and attraction intensity

Abstract: Since community structure is an important feature of complex network, community detection has attracted more and more attention in recent years. Most early researchers focus on identifying disjoint communities, whereas communities in many real networks are overlapped. In this paper, we propose a novel algorithm MCLC with random walks on line graph and attraction intensity to discover overlapping communities. MCLC algorithm first generates a weighted line graph from a undirected network, then divides links into “link communities” through random walks on the line graph. Finally, it transforms the “link communities” to “node communities” using the function of attraction intensity. The “node communities” are permitted overlapped, and the overlapping size is controlled by the threshold of attraction intensity. Experiments on some real world networks validate the effectiveness and efficiency of the proposed algorithm. Comparing overlapping modularity \(Q_{ov} \) with other related algorithms, the results of this algorithm is satisfactory.

KFTrust: P2P trust model based on evaluation rank using Kalman filter

Abstract: Security and privacy have been critically important with the fast expansion of P2P systems, which are open, anonymous and dynamic in nature. Due to such nature of P2P systems, P2P networks present potential threats among nodes. One feasible way to minimise threats is to evaluate the trust values of the interacting nodes. Many trust models have done so, but they fail to properly evaluate the trust values when malicious nodes start to behave in an unpredictable way. To solve the strategically altering behaviour of malicious nodes, this paper designs an evaluation rank-based trust model according to the different recommended trust values. This model computes the trust value by combining the similarity between the nodes and the time decaying features. In the mechanism of finding recommended nodes, a trust-based Kalman filter algorithm is implemented to reduce the system overhead. Finally, in our experiments, we compare our model with other four typical models, the results show that our model is able to defend against malicious nodes' attack effectively and reduce the system overhead.

Finding overlapping communities with random walks on line graph and attraction intensity

Abstract: Since community structure is an important feature of complex network, community detection has attracted more and more attention in recent years. Most early researchers focus on identifying disjoint communities, whereas communities in many real networks are overlapped. In this paper, we propose a novel algorithm MCLC with random walks on line graph and attraction intensity to discover overlapping communities. MCLC algorithm first generates a weighted line graph from a undirected network, then divides links into “link communities” through random walks on the line graph. Finally, it transforms the “link communities” to “node communities” using the function of attraction intensity. The “node communities” are permitted overlapped, and the overlapping size is controlled by the threshold of attraction intensity. Experiments on some real world networks validate the effectiveness and efficiency of the proposed algorithm. Comparing overlapping modularity \(Q_{ov} \) with other related algorithms, the results of this algorithm is satisfactory.

A Weighted Network Model Based on the Correlation Degree between Nodes

Abstract: Many complex networks in practice can be described by weighted network models, and the BBV model is one of the most classical ones. In this paper, by introducing the concept of correlation degree between nodes, a new weighted network model based on the BBV model is proposed. The model takes the both node strength and node correlation into consideration during the network evolution, which better reveals the evolving mechanisms behind various real-world networks. Results from theoretical analysis and numerical simulation have demonstrated the scale-free property and small-world property of the network model, which have been widely observed in many real-world networks. Compared with the BBV model, the added correlation preferential attachment rule in the model leads to a faster network propagation velocity.

Modeling and Analysis MAC Layer Performance for IEEE 802.11s Wireless Mesh Network in Smart Grid

Abstract: IEEE 802.11s based wireless mesh networks (WMN) recently have been proposed as an important networking technology to deploy in Smart Grid (SG) for data collection and remote control purposes. In this paper, we focus on modeling and analyzing the MAC layer performance of EDCA for IEEE 802.11s wireless mesh networks in the smart grid, taking into account the impact of hidden nodes and different QoS requirements of smart grid applications. We first develop a new Markov chain model to analyze the back-off process for different applications with hidden nodes problem. Then based on the analytical model, we derive a few MAC layer performance metrics. Finally, the proposed analytical model is validated via comparing the analytical results with simulation results by ns3 in NAN scenarios with various applications.