Large Social Networks Research Articles

Context-based influence maximization with privacy protection in social networks

As the increase of requirements of accessing and sharing information of people, large social networks have appeared. The influence maximization over social network has been a popular research topic, whose goal is to maximize the expected range of influence by selecting seed nodes to sending information and encouraging nodes in social network to report the messages. On the other side, privacy concerns have became more and more important, both automated and manual efforts are utilized to protection privacy of users. Under the mechanisms of privacy protection, the nodes in social network will not act as they did in the setting of no privacy protection. As far as we know, there are no previous works considering this problem. In this paper, we consider the influence maximization problem with privacy protection mechanisms in social networks.One challenge is that how to abstract the relations between users and information to identify which kinds of information should be protected by the privacy-related mechanisms. A context-based solution is proposed in the paper to face the challenge above and solve the influence maximization problem. First, a context-based information diffusion model (IDC for short) is proposed. Then, the corresponding influence maximization problem (IM-IDC for short) under IDC model is formally defined. Then, the methods about context extraction, influence estimation, and redundant contexts identification are introduced. The IM-IDC problem is shown to be NP-hard, and an efficient approximation algorithm based on greedy strategy is proposed and analyzed theoretically. Finally, experimental results show that our method is efficient.

Dyadic associations reveal clan size and social network structure in the fission–fusion society of spotted hyaenas

AbstractEstimating the size and dynamics of populations is of paramount importance in ecology. In species with uniquely marked individuals, capture–recapture methods can be used to establish population size and to explore associations between individuals. However, very few studies have used cameras traps to focus on group composition in social carnivores, despite being of particular interest in species characterised by “fission–fusion” formation of sub‐groups. Here, we provide estimates of (a) population size, (b) density, (c) clan size, (d) association patterns and (e) social network structure in spotted hyaenas (Crocuta crocuta) based on images from camera traps deployed at waterholes on Ongava Game Reserve (northern Namibia). In a 15 week study period, we identified 32 individuals. Dyadic associations and the resulting social network showed that all but two hyaenas associated directly or indirectly with each other, indicating the presence of one clan of at least 30 individuals, resulting in a density of 8.1 hyaenas/100 km2. We found a very high variability in the tendency of individuals to associate with others. This study confirms a highly dynamic fission–fusion society in spotted hyaenas. We argue that camera traps can provide relevant insights into large carnivore social network structure where associations between individuals are difficult to observe directly.

Legal access to alcohol and academic performance: Who is affected?

Previous research finds that legal access to alcohol hinders the academic performance of college students. However, the existing studies differ materially in magnitudes, suggesting a reduction in subsequent grades of either 0.03 or 0.10 standard deviations. One plausible explanation is that the change in alcohol consumption that occurs upon attaining the minimum legal drinking age (MLDA) differs across student populations. We test this hypothesis by leveraging predictable variation in adherence to the MLDA across students within the same institution. We find that students with limited underage access to alcohol experience the largest declines in academics upon turning 21, while students with large social networks that likely enable underage consumption experience no effect.

Influence Maximization: A Time-Space Efficient Algorithm

Influence maximization is the problem of finding a small subset of nodes (seed nodes) in a social network that could maximize the spread the influence. The algorithm problem of computing influence estimation and influence maximization have been extensively studied for decades. In this paper, we studied the reverse influence sampling method proposed by Byogs et al. in 2014, and some other improved algorithms. We proposed a new algorithm TESA (Time-Space Efficient Algorithm). On large social networks, TESA saves more space than existing algorithms and runs faster. We used data sets tested in the literature to evaluate TESA experimentally and showed that it was superior to the most advanced solutions in terms of running time and memory footprint.

A Novel Description on Edge-Regular q-Rung Picture Fuzzy Graphs with Application

Picture fuzzy model is a generalized structure of intuitionistic fuzzy model in the sense that it not only assigns the membership and nonmembership values in the form of orthopair ( μ , ν ) to an element, but it assigns a triplet ( μ , η , ν ) , where η denotes the neutral degree and the difference π = 1 - ( μ + η + ν ) indicates the degree of refusal. The q-rung picture fuzzy set( q -RPFS) provides a wide formal mathematical sketch in which uncertain and vague conceptual phenomenon can be precisely and rigorously studied because of its distinctive quality of vast representation space of acceptable triplets. This paper discusses some properties including edge regularity, total edge regularity and perfect edge regularity of q-rung picture fuzzy graphs (q-RPFGs). The work introduces and investigates these properties for square q-RPFGs and q-RPF line graphs. Furthermore, this study characterizes how regularity and edge regularity of q-RPFGs structurally relate. In addition, it presents the concept of ego-networks to extract knowledge from large social networks under q-rung picture fuzzy environment with algorithm.

Network Resonance Method: Estimating Network Structure from the Resonance of Oscillation Dynamics

Spectral graph theory, based on the adjacency matrix or the Laplacian matrix that represents the network topology and link weights, provides a useful approach for analyzing network structure. However, in large scale and complex social networks, since it is difficult to completely know the network topology and link weights, we cannot determine the components of these matrices directly. To solve this problem, we propose a method for indirectly determining the Laplacian matrix by estimating its eigenvalues and eigenvectors using the resonance of oscillation dynamics on networks.

RankMerging: a supervised learning-to-rank framework to predict links in large social networks

Uncovering unknown or missing links in social networks is a difficult task because of their sparsity and because links may represent different types of relationships, characterized by different structural patterns. In this paper, we define a simple yet efficient supervised learning-to-rank framework, called RankMerging, which aims at combining information provided by various unsupervised rankings. We illustrate our method on three different kinds of social networks and show that it substantially improves the performances of unsupervised methods of ranking as well as standard supervised combination strategies. We also describe various properties of RankMerging, such as its computational complexity, its robustness to feature selection and parameter estimation and discuss its area of relevance: the prediction of an adjustable number of links on large networks.

A social learning video method: Identifying and sharing successful transformation knowledge for sustainable soil management in Switzerland

AbstractTo enhance sustainable land use, a From Farmer to Farmer project was conducted in Switzerland (2001–2010). A multi‐stakeholder discussion group co‐produced nine videos with experienced farmers and wine producers showing sustainable soil management practices. We analysed the video audio‐visual content and drew on reflections of the co‐production process, referring to concepts of system, target and transformation knowledge, as well as on social learning. The analysis showed a broad range of means (or actions) for sustainable soil management in arable land management, fodder production and wine growing that are aligned to transformation knowledge. The research showed that farmers refer to three phases of social learning, light‐bulb moments, coping with challenges and gaining successful expertise. These are not just linear processes of individuals. Four types of social learning were found in the video analysis: (a) learning from observing actions of others, (b) sharing experiences with storytelling, (c) informal social interactions and (d) being a role model with a large social network. Videos enable transformation knowledge to be shared with peers using storytelling; this powerful narrative communication style provides credibility and respects the ‘thought style’ of the target audience group. We conclude that for successful implementation of sustainable actions, it is important to address a specific target group and share their transformation knowledge built upon system and target knowledge. The social learning video method is a viable way to enable social learning between science, administration and practice and has potential for fostering change in sustainable soil management.

Population size and self-reported characteristics and sexual preferences of men-who-have-sex-with-men (MSM) in Germany based on social network data.

BackgroundIn the absence of detailed information about the population size and behaviour data of men-who-have-sex-with-men (MSM), the estimation of prevalence rates of sexually transmitted infections (STIs) and the design of public health interventions become difficult. The aim of the present study is to estimate the lower boundary of age-specific population sizes and retrieve self-reported information from this population.MethodsWe used publicly accessible data from a large online dating and social network website for MSM in Germany to retrieve data on the age and regional distribution of profiles. The profiles were also stratified by their information on the preferred position during anal intercourse, safer sex, and sexual identity.ResultsA total of 464,873 user profiles correspond to an average 15.2 profiles per 1,000 male inhabitants in Germany, varying between 7.6 and 45.6 across federal states. Although the information on the absolute numbers for different age groups is limited by the search engine, age-specific relative frequencies were found to increase from 12.9 in the age group of 18 to 20 year olds to 24.6 profiles per 1,000 male inhabitants in the 28 to 30 year olds. The data shows age-specific trends for safer sex with an increasing easiness of reporting “never” engaging in safer sex or stating that safer sex “needs discussion” with increasing age. Around one third of profile owners stated to be versatile with respect to the preferred position in anal intercourse. All other options (“only bottom”, “more bottom”, “only top”, “more top”) were preferred equally likely by roughly 10% of profile owners, respectively.ConclusionsOnline social network or dating sites can provide some information about specific populations in the absence of other data sources. The presented results are the first to report age-specific rates of MSM per 1,000 male inhabitants in Germany and may be useful to estimate age-specific prevalence or incidence rates as well as to inform health promotion activities and modelling studies for MSM in Germany.

EXTREME VALUE ANALYSIS WITHOUT THE LARGEST VALUES: WHAT CAN BE DONE?

AbstractIn this paper, we are concerned with the analysis of heavy-tailed data when a portion of the extreme values is unavailable. This research was motivated by an analysis of the degree distributions in a large social network. The degree distributions of such networks tend to have power law behavior in the tails. We focus on the Hill estimator, which plays a starring role in heavy-tailed modeling. The Hill estimator for these data exhibited a smooth and increasing “sample path” as a function of the number of upper order statistics used in constructing the estimator. This behavior became more apparent as we artificially removed more of the upper order statistics. Building on this observation we introduce a new version of the Hill estimator. It is a function of the number of the upper order statistics used in the estimation, but also depends on the number of unavailable extreme values. We establish functional convergence of the normalized Hill estimator to a Gaussian process. An estimation procedure is developed based on the limit theory to estimate the number of missing extremes and extreme value parameters including the tail index and the bias of Hill's estimator. We illustrate how this approach works in both simulations and real data examples.

Identifying key players in large social networks by using a multi-objective artificial bee colony optimization approach

Identifying a set of individuals that have an influential relevance and act as key players is a matter of interest in many real world situations, especially in those related to the Internet. Although several approaches have been proposed in order to identify key players sets, they mainly focus just on the optimization of a single objective. This may lead to a poor performance since the sets identified are not usually able to perform well in real life applications where more objectives of interest are taken into account. Multi-objective optimization seems the natural extension for this task, but there is a lack of this type of methodologies in the scientific literature. An efficient Multi-Objective Artificial Bee Colony (MOABC) algorithm is proposed to address the key players identification problem and is applied in the context of six networks of different dimensions and characteristics. The proposed approach is able to best identify the key players than the ones previously proposed, especially in the context of large social networks. The model performance of the proposed approach has been evaluated according to different quality metrics. The results from the MOABC execution show important improvements with respect to the best multi-objective results in the scientific literature, specifically, in average, 13.20% of improvement in Hypervolume, 120.39% in Coverage Relation and 125.52% in number of non-dominated solutions. Even more, the proposed algorithm is also more robust when repeating executions.

Identifying criminal organizations from their social network structures

Identification of criminal structures within very large social networks is an essential security feat. By identifying such structures, it may be possible to track, neutralize, and terminate the corresponding criminal organizations before they act. We evaluate the effectiveness of three different methods for classifying an unknown network as terrorist, cocaine, or noncriminal. We consider three methods for the identification of network types: evaluating common social network analysis metrics, modeling with a decision tree, and network motif frequency analysis. The empirical results show that these three methods can provide significant improvements in distinguishing all three network types. We show that these methods are viable enough to be used as supporting evidence by security forces in their fight against criminal organizations operating on social networks.

Portal nodes screening for large scale social networks

Network autoregression model (NAM), as a powerful tool to study user social behaviors on large scale social networks, has drawn great attention in recent years. In this paper, we are interested in identifying the influential users (i.e., portal nodes) in a social network under the framework of NAM. Especially, we consider the autoregression model that allows to have a heterogeneous and sparse network effect coefficients. Therefore, the portal nodes take influential powers which are corresponding to the nonzero network effect coefficients. A screening procedure is designed to screen out the portal nodes and the strong screening consistency is established theoretically. A quasi maximum likelihood method is applied to estimate the influential powers. The asymptotic normality of the resulting estimator is established. Further selection procedure is given by taking advantage of the local linear approximation algorithm. Extensive numerical studies are conducted by using a Sina Weibo dataset for illustration purpose.

Parametrization of Spreading Processes Within Complex Networks with the Use of Knowledge Acquired from Network Samples

Information spreading processes are main drivers of viral campaigns. They are usually conducted within large scale social networks. Parametrisation of online campaigns is usually related to allocation of budgets, number of seeds and strategies of their selection. It is hard to predict campaign effects. Proposed in this paper approach uses network samples to select appropriate strategy for use within complete network. Results showed how scaling of samples is affecting approximation quality. Multi-criteria analysis of performance allows for strategy selection according to preferences and goals.

Psychological Gender Express via Mobile Social Network Activities: An Experimental Study on a Gay Network Data

It is well known that the social network users’ behavior can reflect their personal mental profiles. Therefore, in this paper, we conducted an experiment to study users’ gender identity through a mobile social network. Gender identity is the personal sense of one’s own gender, which can be the same or different from their sex assigned at birth, that is, one’s innermost concept of self as male, female, a blend of both, or neither. In general, a person’s psychological gender cannot be observed from the external characteristics of people. And even if some user’s attributes such as physical gender and age can be found in social networks, it is difficult to find out the psychological gender of users. Apart from the fact that it will be hidden as the user’s privacy, the main reason is that users’ behaviors in social networks have no direct relationship with psychological gender. In this paper, we choose a special mobile social network to get around these limitations. Specifically, this paper is based on a real-world large gay social network of more than 10 000 000 users. In this mobile social network, we can indirectly mark the psychological gender of users by their sexual orientation. We collect 1000 users, including their characters, followers, followees and the same data for their followers and followees. Then, we examine an extensive set of features that can be extracted from the data set and train some classifiers to automatically identify the user’s psychological gender. In the special mobile social network, a user’s psychological gender is related to his characteristics, which include his personal characteristics, dating preference, and so on. This paper mainly analyzes users’ personal characteristics, triads, and triadic evolution to infer their psychological gender. The experiments show that all these features are indeed effective in the classification, and even the features considered in previous studies have different implications with gender identity on other social networks. And it turns out that psychological gender as a kind of privacy can be leaked by these unintentional behaviors, and so we give some practical suggestions at the end of this paper to protect users’ privacy.

Who Should Be Invited to My Party: A Size-Constrained k-Core Problem in Social Networks

In this paper, we investigate the problem of a size-constrained k-core group query (SCCGQ) in social networks, taking both user closeness and network topology into consideration. More specifically, SCCGQ intends to find a group of h users that has the highest social closeness while being a k-core. SCCGQ can be widely applied to event planning, task assignment, social analysis, and many other fields. In contrast to existing work on the k-core detection problem, which aims to find a k-core in a social network, SCCGQ not only focuses on k-core detection but also takes size constraints into consideration. Although the conventional k-core detection problem can be solved in linear time, SCCGQ has a higher complexity. To solve the problem of SCCGQ, we propose a Blast Scatter (BS) algorithm, which appoints the query node as the center to begin outward expansions via breadth search. In each outward expansion, BS finds a new center through a greedy strategy and then selects multiple neighbors of the center. To speed up the BS algorithm, we propose an advanced search algorithm, called Bounded Extension (BE). Specifically, BE combines an effective social distance pruning strategy and a tight upper bound of social closeness to prune the search space considerably. In addition, we propose an offline social-aware index to accelerate the query processing. Finally, our experimental results demonstrate the efficiency and effectiveness of our proposed algorithms on large real-world social networks.

ERATO: Trading Noisy Aggregate Statistics over Private Correlated Data

With the commoditization of personal privacy, pricing private data has become an intriguing problem. In this paper, we study noisy aggregate statistics trading from the perspective of a data broker in data markets. We thus propose ERATO, which enables aggr E gate statistics p R icing over priv AT e c O rrelated data. On one hand, ERATO guarantees arbitrage freeness against cunning data consumers. On the other hand, ERATO compensates data owners for their privacy losses using both bottom-up and top-down designs. We further apply ERATO to three practical aggregate statistics, namely weighted sum, probability distribution fitting, and degree distribution, and extensively evaluate their performances on MovieLens dataset, 2009 RECS dataset, and two SNAP large social network datasets, respectively. Our analysis and evaluation results reveal that ERATO well balances utility and privacy, achieves arbitrage freeness, and compensates data owners more fairly than differential privacy based approaches.

Cost-efficient viral marketing in online social networks

Word-of-mouth plays an important role in the spread of influence by means of a cascading manner, e.g., the promotion of new products, information dissemination, etc. Today’s Online Social Networks (OSNs) provide viral marketing for many companies to conduct business promotion, and the idea behind viral marketing comes from the word-of-mouth effects. Given a predefined budget (e.g., the set size k), influence maximization selects a set of initial users to help companies to promote their products. The budget-making often is of challenging owning to the hardness of trade-off between profit and cost. Most of recent research on influence maximization primarily study the problem of independent product promotion, but few take into account the impacts of other existing products that users have already owned. Moreover, the current solutions encounter scalability problem whenever facing a large scale social network. In this paper, we explore the viral marketing by defining a general influence maximization problem where a practical scenario of coexistence of multiple products is considered. To capture the impacts of other existing products, we put forward a new method to measure the influence between users. Different from previous work, the goal of our general problem is to maximize the profit/cost ratio which can avoid the difficulty of budget-making and reflects maximization effects better. Then, we propose a ($\frac {1}{2} + \varepsilon $)-approximate algorithm with linear complexity to solve the problem, and further design a distributed implementation to achieve good scalability. Extensive experiments on real world datasets validate the effectiveness and efficiency of our solution.

Evolution of corporate reputation during an evolving controversy

PurposeThe purpose of this paper is to investigate the evolution of online sentiments toward a company (i.e. Chipotle) during a crisis, and the effects of corporate apology on those sentiments.Design/methodology/approachUsing a very large data set of tweets (i.e. over 2.6m) about Company A’s food poisoning case (2015–2016). This case was selected because it is widely known, drew attention from various stakeholders and had many dynamics (e.g. multiple outbreaks, and across different locations). This study employed a supervised machine learning approach. Its sentiment polarity classification and relevance classification consisted of five steps: sampling, labeling, tokenization, augmentation of semantic representation, and the training of supervised classifiers for relevance and sentiment prediction.FindingsThe findings show that: the overall sentiment of tweets specific to the crisis was neutral; promotions and marketing communication may not be effective in converting negative sentiments to positive sentiments; a corporate crisis drew public attention and sparked public discussion on social media; while corporate apologies had a positive effect on sentiments, the effect did not last long, as the apologies did not remove public concerns about food safety; and some Twitter users exerted a significant influence on online sentiments through their popular tweets, which were heavily retweeted among Twitter users.Research limitations/implicationsEven with multiple training sessions and the use of a voting procedure (i.e. when there was a discrepancy in the coding of a tweet), there were some tweets that could not be accurately coded for sentiment. Aspect-based sentiment analysis and deep learning algorithms can be used to address this limitation in future research. This analysis of the impact of Chipotle’s apologies on sentiment did not test for a direct relationship. Future research could use manual coding to include only specific responses to the corporate apology. There was a delay between the time social media users received the news and the time they responded to it. Time delay poses a challenge to the sentiment analysis of Twitter data, as it is difficult to interpret which peak corresponds with which incident/s. This study focused solely on Twitter, which is just one of several social media sites that had content about the crisis.Practical implicationsFirst, companies should use social media as official corporate news channels and frequently update them with any developments about the crisis, and use them proactively. Second, companies in crisis should refrain from marketing efforts. Instead, they should focus on resolving the issue at hand and not attempt to regain a favorable relationship with stakeholders right away. Third, companies can leverage video, images and humor, as well as individuals with large online social networks to increase the reach and diffusion of their messages.Originality/valueThis study is among the first to empirically investigate the dynamics of corporate reputation as it evolves during a crisis as well as the effects of corporate apology on online sentiments. It is also one of the few studies that employs sentiment analysis using a supervised machine learning method in the area of corporate reputation and communication management. In addition, it offers valuable insights to both researchers and practitioners who wish to utilize big data to understand the online perceptions and behaviors of stakeholders during a corporate crisis.

How to maximize advertising performance in online social networks

The influence maximization problem is widely studied, but previous studies have assumed that the cost to activate each seed node was identical. In this paper, we consider different activation costs and investigate a new problem: the budget-aware influence maximization problem (BIM). This problem is NP-hard, which motivates our interest in its approximation. We develop two greedy algorithms for BIM, namely, BG and GMUI, and show that GMUI obtains a solution that is provably . Then, we consider the average profit to activate a common user and introduce GMUN, an improved GMUI algorithm. Finally, we evaluate our algorithms with experiments using two large real social networks. The results show that GMUI performs best in terms of influence, whereas GMUN creates leverage between the budget and net income, indicating that the marginal net income/costs ratio for each seed node selected by GMUN can satisfy enterprises.