Given a graph `G', Influence Maximization is the problem of finding a subset of nodes of size `k' that would maximize the spread of influence in G. This problem has applications in viral marketing studies and spread of information through `word of mouth'. The problem, as defined by Domingos and Richardson, can be stated as follows: If we can give a product to a small subset of the population such that these people will convince the most number of people to adopt the product in the future, which subset would we choose? Discount heuristics provide great computational speed up in comparison to the traditional greedy algorithm, which runs for hours for networks with tens of thousands of nodes. In this work, we cite a perceived limitation in the degree discount heuristic for Influence maximization, and develop three new discount heuristics, namely Closeness discount, Betweenness discount and PageRank discount for comparison against the degree discount. We show that using degree discount heuristic still leads to the best seed set selection and hence show that the perceived limitation in the degree discount heuristic does not exist. In addition, we also show that PageRank discount beats Degree Discount in terms of Influence Spread when heterogeneous probabilities are used, thus showing that merely considering graph characteristics without taking into account other nodal properties is insufficient.
More real-world complex systems contain mass of data and are composed of not only a single layer network. A fundamental problem on these systems is how to get maximum information entropy but economizing energy. Here we explore this problem on a two layer asymmetry-coupled network. On the assumption that a node could couple with more than one counterparts on the other layer network, we found for a given average degree, the entropy of the system varies differently according to the influence coefficient of asymmetry depending on the number of node's counterparts. We also show that the energy expanded increase if nodes have fewer counterparts or in a not obvious community structure of each layer network, which give a clue to maximizing entropy while consuming least energy on this two layer asymmetry-coupled network.
Heterogeneous networks have become a commonly used model to represent complex and abstract social phenomena. They allow objects to have many different relationships and represent relationships by semantic paths which connect object types via a sequence of relations. A major challenge in community detection on heterogeneous networks is how to organize and combine different semantic paths. In order to acquire desired clustering, we propose a novel community detection method for heterogeneous networks based on matrix decomposition and semantic paths. The major advantage of this method is to treat objects individually and to assign them with different combinations of semantic-path weights so as to improve the clustering quality. The comparative experiments of the proposed method with another two state-of-the-art methods, spectral clustering and path-selection clustering, confirms that it can acquire desired clustering results better.
Protein structure prediction is an important area of research in bioinformatics. In this paper, we select the features of correlation coefficient sequence and special amino acid composition. The support vector machine and a particular framework of ECOC are employed as classification model. To evaluate the efficiency of the proposed method we choose three benchmark protein sequence datasets (25PDB, 40PDB and ASTRAL) as the test dataset. The final results show that our method is efficient for protein structure prediction.
The model we propose in this paper takes its roots from the OSI (ISO 1984) model, the TCP/IP model (US DoD4 1970), and the Internet model, but it puts its focus on Internet of Things (IoT) specific features and issues. All the previous models have a great value, going beyond any discussion, but simply they have not been conceived with the IoT issues and features in mind. IoT may need more than a computer network communication model! We developed a new model compatible with the prospects of IoT. It should also be noted that there are efforts in recent years to produce a new reference model for communications to keep pace with the world of Internet for everything. The most prominent of these attempts are produced by the IoT Architecture project (IoT-A 2013), which seems appropriate, but we further propose additions and modifications in this paper.
The Social and Smart (SandS) project ecosystem is compounded of household appliance users sharing recipes for the used of appliances, an intermediate control layer, and an intelligent social layer which aims to optimize the appliance recipes maximizing user satisfaction. We consider two aspects of the social intelligence, the innovation producing new recipes for unkown user tasks, and the adaptation to personalize the recipe to an individual user on the basis of his/her specific feedback. The second aspect is proposed to be dealt with by Reinforcement Learning approach, thus user feedback becomes the system reward. In this paper we discuss such an architecture based on the actor-critic approach, providing some experimental results on synthetic datasets that demonstrate the feasibility of the approach, previous to real life implementations.
Researchers regularly access and review large amounts of literatures. In the previous work, we presented a bookmarklet-triggered literature sharing system, which combines bibliography functionalities along with DOI content negotiation services. In this paper, we have made secondary development work to integrate literature recommendation functionalities into this system. We introduce a hybrid approach in parallel to recommend related articles to researchers. First, we collect a large amount of published and new articles using crawlers and RSS listeners to address cold start issue. Second, we adopt Latent Dirichlet Allocation (LDA) as the topic model to category literatures. For one kind of literatures related to researchers' interest, we use collaborative filtering techniques to make further analysis based on implicit user feedbacks in this system. Finally, we take matrix factorization with Alternating Least Squares (ALS) in parallel to compute the top-N recommendations per user.
The use of Centrality Measures (CMs) in the analysis of Online Social Networks (OSNs) has proved to be an effective strategy for identification of potentially influential users who can disseminate information on the network faster and more efficiently. Nevertheless, the selection based on individual CMs focuses in a particular user's attribute that singly may not reflect its real importance. In this sense, this paper presents a multicriterial approach for analyzing centrality in OSNs by using the Analytic Network Process (ANP) method, which are modeled by means of the interrelationships between CMs to provide greater robustness in the central user's selection. A set of simulations was also performed, showing the consistency and good performance of the proposed method.
In the last decade, social networks have increasingly attracting the attention of several researchers in various fields. By tradition, social network analysis (SNA) is performed on static graphs but this representation is very limited for a sound, network analysis. Thus, thanks to the availability of large social networks data sets, the interest in modeling how these networks evolve dynamically has increased steadily. Many studies have been oriented toward understanding the behavior and the evolution of network structures over time. This paper does the survey of complex networks models and methods which are proposed to reproduce structural changes of these graphs.
Community identification in large networks is one of the most popular Social Network Analysis applications, and many algorithms have been proposed. The visualisation of the identified structure remains a problem in large networks. The traditional graph-based visualisation does not scale well with many communities and their numerous relations among each other. In this paper, we propose a visualisation based on abstracted adjacency matrices, which scales much better, since there are no overlaps in the two-dimensional matrix. We also propose a couple of enhancements and tweaks to get the best possible user experience with this approach.
This paper intends to extend the possibilities available to researchers for the evaluation of directed networks with the use of randomly generated graphs. The direct generation of a simple network with a prescribed degree sequence still seems to be an open issue, since the prominent configuration model usually does not realise the degree distribution exactly. We propose such an algorithm using a heuristic for node prioritisation. We demonstrate that the algorithm samples approximately uniformly. In comparison to the switching Markov Chain Monte Carlo algorithms, the direct generation of edges allows an easy modification of the linking behaviour in the random graph, introducing for example degree correlations, mixing patterns or community structure. That way, more specific random graphs can be generated (non-uniformly) in order to test hypotheses on the question, whether specific network features are due to a specific linking behaviour only. Or it can be used to generate series of synthetic benchmark networks with a specific community structure, including hierarchies and overlaps.
Information systems support and ensure the practical running of most critical business processes. There exists or can be reconstructed a record (log) of the process running in the information system with information about the participants and the processed objects for most of the processes. This research was realized in the environment of the enterprise information system SAP. Participants of business processes stand in different relationships. We are interested in the relationships that are not explicitly seen from the process logs, but which are detectable by research methods of social networks and communities in social networks. Our work constructs the social network from the process log in the given context and then it finds communities in this network. Found communities were analyzed using knowledge of the business process and the environment in which the process operates. We found that identified communities have reasonable representation in the actual process, and this opened up a new dimension of knowledge that can be analyzed from the process log. This approach seems to be promising for detailed analysis.
This paper describes functions of a system designed for the behavior analysis of e-commerce clients. It enables user identification and client behavior extraction for interacting with web site customers. General approaches used in the field of Web Usage Mining are presented together with proposals to extend the data base with the information gained from e-commerce site forums and queries. Our system carries out an evaluation and rating of opinions, and our approach is based on linguistic and the statistic treatment of natural language. Three different methods for classifying opinions from clients' forum are used, and two new methods, based on linguistic knowledge to assign a mark dependent upon the client's emotions and opinions described in forum comments, have been introduced.
Most computational techniques that analyze Online Social Networks (OSNs) aim to discover patterns in a network's structure and the behavior of its users, but do not seek to understand how people's motives lead to these patterns. Studying the social effects that cause these patterns, however, can produce deeper insights that may transcend a specific network and are generically applicable. Therefore, a more promising approach is to anchor computational techniques to the underlying social effects that can explain the reasons behind why users interact the way they do. In this paper, we discover how the social effects of stature, relationship strength, and egocentricity shape the interactions among Facebook users. These effects are explored through transitivity in triads, which are network units that capture dynamics among triples of users. The analysis suggests that Facebook interactions are influenced by users with concentrated stature and strong bonds. However, the activities of popular and over-active users have little influence.
The paper presents several novel ideas on how to understand social interaction as an intelligent computing phenomena, proposing a taxonomy of social systems regarding intelligent behavior that may be useful to set the stage. A key distinction between unconscious and subconscious computing is drawn. Two instances of systems, which are in diverse stages of development, showing subconscious social intelligent computing are discussed.
Cloud Computing is becoming a promising technology for processing a huge chunk of data. Hence, its security aspect has drawn the attentions of researchers and academician. The security of the cloud environment must be reliable as well as scalable.The cloud environment is vulnerable to many security attacks. Attacks can be launched individually or in tandem. In this article, the overview of port-scan attack and the response of IDS are studied. The experimentation is carried out using virtual-box and SNORT, the open-source IDS.
Online social networks like Weibo and Twitter consist of billions of users and connections, and traditional approaches which are based on serial algorithms and leveraged only a single node or even a single core cannot suffice the that scale of data any more. We propose new distributed quasi-parallel breadth-first search scheme, the common graph traversal algorithm, based on the MapReduce framework, which has better performance (up to one scale of magnitude less time complexity for single-source cases or even better for multiple-source cases) than Pegasus, the state-of-the-art graph mining library, in terms of the complexity of computation and the I/O load. We apply our algorithms on the Weibo dataset, crawled from its website, which contains 135 million users and 10.2 billion directed connections among them, and occupies up to 400 gigabytes. The dataset is by far the largest one of online social networks in research. Based on the Weibo dataset with extremely skewed degree distribution, we give the empirical time complexity and I/O load analysis in each iteration of our proposed methods. Also, We ran the experiments on a 20-node Hadoop cluster to validate our analysis, and the results conform to our predicted empirical results.
Visualization is an important part of network analysis. It helps find features of the network that are not easily identifiable. Visualization dynamics of the network is very useful. Evolution of network and communities during time can help us understand social mechanisms behind the network. Visualization of the dynamics is not an easy task. There are several issues that have to be solved for correct visualization. We present our approach to the visualization of weighted networks based on Sammon's projection and linear approximation. Our goal in this paper is to introduce a method for visualizing dynamics of the social network. Results are illustrated by several 3D layout snapshots of the co-authorship network extracted from the DBLP database.
Social Network Services have become an important medium for people to communicate ideas and share interests in recent years. Blogs published and shared by users in this virtual world are one of the main sources of user-generated information. Classifying these freestyle blogs can help understand user interests and assist applications such as search and marketing. In this paper, we propose a new method of multi-label classification for Chinese blogs. By applying Dempster-Shafer theory on semantic word similarity algorithms, we achieve automatic classification without use of difficult-to-obtain training sets. Experiments were conducted on real world data from RENREN.com, the biggest SNS (Social Network Services) in China. Results show that the proposed method achieves satisfactory performance in multi-labeling real world SNS blogs as well as corpus.
Evolution of cooperative strategies was examined for spatial iterated prisoner's dilemma (IPD) games in many studies. Lattices and networks were frequently used as spatial structures where a single player was assigned to each node. It was demonstrated that spatial structures were beneficial for the evolution of cooperation. In this paper, we examine the effect of the choice of a network structure on the evolution of cooperation in a network-based spatial IPD game. We use a variety of networks. They are different from each other in the network size and the number of edges from each node. Some networks have edges between randomly selected nodes while edges of other networks are only between adjacent nodes. Memory-based lookup tables are used as strategies of players. Computational experiments are performed under two settings with respect to the noise in action selection. One is a noise-free setting where each player always chooses the suggested action by its strategy. The other is a noisy setting where each player chooses a different action from the suggested one with a pre-specified error probability. We examine the effects of the network size, the number of edges, the number of opponents and the addition of randomly specified edges as well as the memory length on the evolution of cooperative strategies.
