PaperlayerView exact revision on arXiv

Source-linked AI summary

The Four Dimensions of Social Network Analysis: An Overview of Research Methods, Applications, and Software Tools

David Camacho, Àngel Panizo-LLedot, Gema Bello-Orgaz, Antonio Gonzalez-Pardo, Erik Cambria

arXiv:2002.09485v1cs.SIcs.CYcs.LG

TL;DR

Social network analysis has expanded rapidly, creating a need to understand its research landscape and the capabilities of available technologies. The paper reviews the field, defines four evaluation dimensions and related metrics, and assesses SNA tools and frameworks. It finds that most analyzed tools handle graphs of around 100,000 nodes, while only a few support Big Data and those tools show low or medium average Knowledge Discovery capabilities.

  • Problem

    Rapidly growing social-network data and applications create challenges for extracting and analyzing knowledge across diverse network settings.

  • Method

    The paper combines an updated SNA literature review, a scientometric study, four evaluation dimensions, and quantitative assessment of 20 software tools and frameworks.

  • Results

    Most analyzed tools handle graphs of around 100,000 nodes, but few support Big Data and those tools have low or medium average Knowledge Discovery capabilities.

  • Takeaways & Limitations

    The four dimensions—Pattern & Knowledge Discovery, Information Fusion & Integration, Scalability, and Visualization—support quantitative assessment of SNA technology maturity.

  • Takeaways & Limitations

    The proposed dimensions and metrics are an initial, incomplete framework rather than the only possible evaluation scheme.

Abstract

from arXiv · show

Social network based applications have experienced exponential growth in recent years. One of the reasons for this rise is that this application domain offers a particularly fertile place to test and develop the most advanced computational techniques to extract valuable information from the Web. The main contribution of this work is three-fold: (1) we provide an up-to-date literature review of the state of the art on social network analysis (SNA);(2) we propose a set of new metrics based on four essential features (or dimensions) in SNA; (3) finally, we provide a quantitative analysis of a set of popular SNA tools and frameworks. We have also performed a scientometric study to detect the most active research areas and application domains in this area. This work proposes the definition of four different dimensions, namely Pattern & Knowledge discovery, Information Fusion & Integration, Scalability, and Visualization, which are used to define a set of new metrics (termed degrees) in order to evaluate the different software tools and frameworks of SNA (a set of 20 SNA-software tools are analyzed and ranked following previous metrics). These dimensions, together with the defined degrees, allow evaluating and measure the maturity of social network technologies, looking for both a quantitative assessment of them, as to shed light to the challenges and future trends in this active area.

1 Introduction

The paper reviews social network analysis across a rapidly expanding, multidisciplinary research landscape and introduces four dimensions and related metrics for assessing SNA technologies.

  • Exponential OSN growth has generated broad interest in network structure, evolution, information diffusion, and extractable patterns across several disciplines.
  • The authors use scientometric analysis to select active SNA research areas, application domains, and emerging topics for review.Selected areas include graph theory, community detection, information diffusion, user profiling, topic extraction, and sentiment analysis; application domains include healthcare, marketing, tourism, and cybersecurity.
  • The paper defines four SNA dimensions—Pattern & Knowledge discovery, Information Fusion & Integration, Scalability, and Visualization—to assess tools and frameworks.These dimensions correspond to discovering patterns, integrating heterogeneous sources, processing large networks, and representing network information.
  • A global Capability metric, CSNA, combines the proposed metrics to rank SNA technologies, frameworks, and tools.
  • The assessment analyzes 20 SNA tools and frameworks to identify their strengths and weaknesses and guide researchers in selecting technologies.The assessment includes relevant software tools and frameworks available on the Internet.

2 SNA Scientometric Analysis

The scientometric analysis traces SNA’s historical influences, identifies prominent topics and application domains, and highlights emerging research areas. It combines RPYS, text analysis, and network-oriented methods to characterize the field.

  • Historical influence: RPYS identifies pronounced citation peaks in 1967, 1973, 1977, 1979, 1988, 2011, and 2012.Earlier influential works appeared mainly in Social Science journals, while the most recent high-impact articles appeared in Computer Science journals.
  • Historical influence: The analysis links earlier SNA research to diffusion, mobility, community organization, interpersonal links, and structural centrality measures.These findings describe foundational contributions from the 1970s and 1980s.
  • Topics and applications: LDA-based textual analysis processes article keywords to identify relevant topics and visualize frequent terms in a word cloud.The keyword analysis especially associates SNA with Pattern & Knowledge Discovery and applications such as healthcare.
  • Topics and applications: Recent SNA research centers on Data Science, Big Data, network analysis, social media, sentiment analysis, text mining, and information diffusion.Prominent application areas include health, marketing and business, and tourism.
  • Emerging areas: Fake news, misinformation, and cybersecurity emerged as increasingly active areas despite not being directly highlighted by the scientometric analysis.The paper treats these topics as emerging areas because of their potential relevance in the near future.

3 Techniques and algorithms

The paper situates SNA techniques within a broad multidisciplinary landscape. From a Computer Science perspective, it emphasizes algorithms for processing, extracting, and visualizing knowledge from large networks.

  • Research landscape: SNA spans multidisciplinary fields ranging from Social Sciences to Science and Engineering.The section focuses specifically on current findings from the Computer Science perspective.
  • Computational techniques: Computer Science research in SNA develops algorithms to process large amounts of network knowledge.The passage connects this work with Data Science and Big Data.
  • Computational techniques: These algorithms address knowledge extraction and visualization as well as network processing.The passage frames all three activities as closely associated with Data Science and Big Data.

3.1 Structural-based Analysis

Structural-based analysis represents social networks as graphs and studies their topology, models, metrics, communities, dynamics, and information propagation. These approaches support both network characterization and the analysis of large-scale social phenomena.

  • Graph representation: SNA represents social entities as graph nodes connected by edges that encode relationships.Graphs may be directed or undirected, and edges can carry numerical weights.
  • Graph models: Graph and network models reproduce characteristics of real networks using measures such as degree distributions, path lengths, and clustering.Real-world social networks commonly exhibit short distances, relatively high transitivity, and power-law degree distributions.
  • Network metrics: Centrality measures assess node relevance and can identify influential people in online social networks.Degree centrality ranks highly connected nodes, while eigenvector centrality incorporates the importance of neighboring nodes.
  • Community detection: Community detection includes non-overlapping methods that assign each node to one group and overlapping methods that allow membership in several groups.Overlapping approaches better represent cases such as people belonging to multiple online communities.
  • Community dynamics: Dynamic community detection models community evolution over time, but snapshot granularity and algorithmic instability can affect interpretation.Temporal models provide greater expressivity while requiring more complex analysis and linking communities across adjacent snapshots.
  • Information diffusion: The rapid propagation of content through highly connected networks produces viral content that can support popularity-seeking and marketing campaigns.The passage connects network connectivity with fast reach among large numbers of potential readers.

3.2 Content-based Analysis

Content-based analysis examines interaction content and user activity to extract topics, opinions, profiles, interests, and influence from online social networks. It combines language-oriented methods with network structure to support classification, profiling, recommendation, and trend analysis.

  • Content-based SNA: Content-based SNA analyzes interactions to extract topics or opinions and uses user behavior to construct profiles.User profiling can use clustering, behavioral analysis, content analysis, and face detection over aggregated activity data.
  • Communication-based profiling: Email-based profiling methods map message keywords to recipients and use communication behavior to rank or predict similar users.One implementation filters out infrequent email communications below a threshold.
  • Influence analysis: Influential-blogger methods use comments, interests, links, reproduction patterns, productivity, and interpersonal similarity to identify or recommend influential users.These approaches can also characterize temporal and behavioral blogger patterns.
  • Topic extraction: Topic extraction identifies abstract topics in document collections using methods such as mixture models, latent semantic indexing, LDA, and knowledge-driven approaches.Combined with SNA, these methods can mine user interests in online social networks.
  • Topic and community models: ART, RART, and CART models combine authors, recipients, roles, topics, and communities to analyze text-based networks.CART was used to extract email communities according to the topics covered by network members.
  • User profiling: User-interest systems classify generated content and rank interests using topical representations, including multilingual Twitter-interest classification.One approach computes graph distance between represented interests and topical resources.
  • Sentiment analysis: Sentiment analysis processes social-media content across multiple platforms and supports applications including healthcare, political forecasting, tourism, rumor detection, and dialogue systems.The passage characterizes sentiment analysis as a complex NLP problem rather than only a categorization task.
  • Sentiment and network structure: Combining sentiment features with social-network structure can enhance classification in complex online environments.Applications also include trend discovery, user profiling, influencer detection, and polarization analysis.

4 Application domains

The paper selects application domains through a scientometric study, focusing on Healthcare, Marketing, Tourism and Hospitality, and Cybersecurity, while also examining Politics and fake-news and misinformation detection as emerging areas.

  • Selected domains: The scientometric study selected Healthcare, Marketing, Tourism and Hospitality, and Cybersecurity as application domains for detailed analysis.The paper also identifies Politics and detection of fake news and misinformation as emerging areas of SNA.

4.1 Healthcare

Healthcare applications of online social-network analysis examine social interaction, health-related behavior, peer influence, and public-health signals. The reviewed work spans propagation modeling, risk behavior, and surveillance-oriented analysis.

  • Healthcare interactions: Social relationships affect physical and mental health, while online networks create interaction possibilities beyond the traditional doctor-to-patient paradigm.Patients use online social networks to share experiences and disease-related information.
  • Health-behavior propagation: Propagation and virality models analyze health-related behaviors, including obesity spread through a densely interconnected network of 12,067 people observed from 1971 to 2003.The cited study used repeated measurements over a 32-year period.
  • Peer influence: Adolescent tobacco and alcohol use is associated with friends’ use, motivating analysis of online and offline friendship networks and peer influence.Online activities may broker peer influence by encouraging or hindering the effects of risk behaviors.
  • Public-health surveillance: Online social-network text enables real-time analysis of mood and social behavior patterns for identifying, assessing, and verifying potential public-health risks.Such analyses can support timely dissemination of public-health alerts.

4.2 Marketing

Social media supports brand promotion, customer feedback, electronic word-of-mouth, and customer relationship management. Studies use social media data to examine brand sentiment and forecast outcomes.

  • Companies use social media to promote products and assess customers’ opinions of brands and products.
  • Electronic word-of-mouth lets customers share opinions about products and brands with many other customers.Twitter studies found that 19% of analyzed tweets mentioned a brand, and 20% of those mentions expressed brand sentiment.
  • Social media data can reveal customer perceptions and purchasing decisions related to brands and competitors.
  • A Twitter model based on tweet-creation rate outperformed market-based predictors when forecasting movie box-office revenues.
  • Firms also use social media to support customer experience and customer relationship management through increased interaction.

4.3 Tourism & Hospitality

Tourism and hospitality research uses social media to study marketing, tourist decisions, satisfaction, experiences, and competitive position. Analyses draw on textual posts and photographs, with methods determined by the data source.

  • Tourism destinations and hotels use social media for marketing, and half of tourists change travel plans after researching trips on social media.
  • Tourism studies analyze textual posts and photographs as distinct social-media data sources.
  • Textual user-generated content reveals tourist opinions, experiences, satisfaction, and dissatisfaction with tourism products.
  • Research links tourism social-media data to tourist satisfaction attributes, guest experience, competitive position, and hotel electronic word-of-mouth.
  • Photo metadata, including geotags, supports analyses of tourist behavior, travel-path recommendations, and image elements viewed by users.

4.4 Cyber Security

Cybersecurity applications of SNA address crime analysis and counter-terrorism by examining communication, geographic patterns, network structure, and terrorist language. These approaches support detection and prediction but can face computational limits.

  • SNA supports law-enforcement strategies against cyber-crime and cyber-terrorism by analyzing rapidly propagating messages and network connections.
  • Crime-analysis systems combine natural-language processing and similarity methods to facilitate communication between citizens and government agencies.
  • Geographic knowledge-discovery techniques identify co-distribution patterns that may help predict where, when, and why crimes are likely to occur.
  • Hotspot mapping uses past crime events and geographic point distributions to identify regions where future crimes may occur.
  • Counter-terrorism research examines how terrorist organizations use social networks to promote ideology and recruit individuals.
  • Four counter-terrorism application areas include identifying terrorist language and linguistic markers in social-network text.
  • Betweenness centrality can identify critical network nodes, but its O(nm) computational cost motivates heuristic and bio-inspired alternatives.

4.5 Emerging Areas

Emerging SNA applications include politics, misinformation and fake-news detection, and multimedia analysis. Research combines content, network, and visual methods to study polarization, diffusion, hate speech, and increasingly diverse social-media content.

  • Emerging SNA areas address politics, fake news and misinformation, and multimedia-information integration.
  • Political social-media research examines hate speech, political polarization, community finding, and information diffusion.
  • A study of 55,000 German right-wing hate tweets reported 84% accuracy for detecting right-wing hate speech.
  • Analysis of 1,150,000 messages from about 220,000 users found retweet structures clustered by political party and party-related differences in language.
  • A political-leaning classifier achieved 86% accuracy, while another study found Twitter use likely reinforced partisan loyalties during elections.
  • Political information-diffusion studies examine partisan homophily, emotional language, retweet rates, and echo-chamber communication.
  • Fake-news detection approaches divide into linguistic methods analyzing writing features and network methods analyzing interactions among users and posts.
  • Although text remains the most commonly analyzed social-media content, improved connectivity and smartphones have increased interest in multimedia analysis.

5 Discussion on Research Methods and Application Domains

Social network analysis must address massive data, multidisciplinary collaboration, knowledge extraction, visualization, and technology maturity. The paper organizes these challenges into four dimensions and proposes quantitative tools to assess SNA technologies and identify future research opportunities.

  • Billions of users generate huge amounts of social-network data, exceeding classical processing capabilities even when only a small portion is analyzed.
  • Structural- and content-based algorithms address knowledge discovery, while scalable methods are needed to manage exponentially growing data.
  • The paper proposes four dimensions for assessing the maturity of technologies currently available for online social networks.
  • Spider graphs and the CSNA metric provide a quantitative measure of SNA tools and frameworks.
  • The proposed dimensions support investigation of future trends, challenges, improvement areas, and emerging technologies with potential impact on science, industry, and society.

6 The Four Dimensions of Social Network Analysis

The paper defines four SNA dimensions—Pattern & Knowledge Discovery, Information Fusion & Integration, Scalability, and Visualization—and operationalizes them through quantitative degrees and a global capability metric. These measures assess how tools handle network knowledge, data diversity, scale, and visual representation.

  • Scalability: Scalability evaluates whether methods maintain service quality as datasets or problem complexity grow, using processable nodes and edges as key indicators.The proposed Space-Time degree assigns Low, Medium, or Large values according to the maximum network size processed without performance degradation.
  • Dimensions: The four dimensions measure pattern discovery, information integration, scalability, and visualization in SNA tools and frameworks.They are inspired by Big Data V-models and support assessment of technology maturity and readiness.
  • Information Fusion & Integration: Information Fusion & Integration measures the diversity of data that SNA methods can combine, including multiple sources and increasingly complex representations.Multi-representation ranges from basic unweighted graphs to weighted graphs and advanced multi-layer graphs.
  • Visualization: Visualization assesses how tools represent network information and support interaction through visual variables such as zoom, filtering, and highlighting.The visualization degree combines the number of visual variables and available interactions using weighted characteristics.
  • Visualization: Node-link diagrams can become unreadable for small-world networks, motivating alternative representations and visualization measures.The paper identifies the resulting dense display as a major SNA visualization challenge because OSNs are usually small-world networks.
  • Global capability metric: The SNAdegrees map the four research questions to quantitative dimension measures, while CSNA aggregates them into a capability value for ranking technologies.CSNA is based on the area of the polygon formed by the normalized dimension scores; the paper notes that comparisons across different technology types may lack meaning.

7 Frameworks & Tools Analysis

The analysis evaluates 20 SNA-software tools using global and dimension-specific metrics, separating license categories and examining capability distributions and relationships. Graphistry leads the global ranking, while different tools stand out across dimensions and no tool dominates every dimension.

  • Tool selection and evaluation: 20 SNA-software tools were selected from an initial list of 70 candidates and evaluated using a top-down process beginning with the global capability metric.The candidates included computing libraries, web applications, distributed applications, and desktop applications, with both open-source and proprietary licenses represented.
  • Global capability ranking: The analysis ranks tools under Proprietary or Open-Source and Only Open-Source licenses, with Graphistry achieving the best global score.Graphistry has the highest possible scores in Information Fusion, Scalability, and Visualization, but its Knowledge Discovery score is below average and ranks 13th in that category.
  • Dimension-specific results: ORA-LITE/PRO leads Knowledge Discovery and is the only tool scoring above zero on every feature in that dimension, while Graphistry leads Information Fusion and reaches maximum ratings across all its features.Knowledge Discovery includes sparse coverage of Opinion Mining, Homophily, and Virality; Multi-Modality is similarly sparse within Information Fusion.
  • Dimension-specific results: Visualization is led by ORA-LITE/PRO, Graphistry, and Neo4j, followed by Gephi and Cytoscape, while Measures Topology, Link analysis, and Functional features score highly across tools.Opinion Mining, Homophily, and Multi-Modality are addressed by comparatively few tools, producing uneven feature distributions across dimensions.
  • Dimension relationships: Visualization relates proportionally to Knowledge Discovery and Information Fusion, whereas Information Fusion and Knowledge Discovery lack the hypothesized proportional relation.The hypothesized inverse relations involving Scalability and Knowledge Discovery or Information Fusion are reported as present; Figure 12 uses pairing plots, averages, density, and Pareto fronts to inspect these relations.

8 Conclusion, Future Trends and Challenges

The conclusion frames SNA as a rapidly expanding field whose scale makes tool and algorithm selection difficult, and evaluates technologies through four dimensions and related metrics. The assessment identifies uneven capabilities, recurring gaps, and future priorities while noting that the proposed framework is provisional.

  • Motivation: The field’s exponential growth makes it difficult for researchers to analyze its state of the art and select appropriate tools or algorithms.The paper addresses this challenge through a structured review and quantitative assessment.
  • Evaluation framework: The paper evaluates SNA technologies using Pattern & Knowledge discovery, Scalability, Information Fusion & Integration, and Visualization dimensions with related degrees.A quantitative evaluation covers 20 popular SNA software tools.
  • Pattern & Knowledge discovery: Topology measures, link analysis, and static community detection are common, whereas dynamic community detection, opinion mining, virality, and homophily are relatively rare.The authors suggest that OSN content is not being fully exploited by current tools.
  • Scalability: Most analyzed tools handle graphs of around 100.000 nodes, but few support BigData, and those that do have low or medium average Knowledge Discovery capabilities.The paper anticipates growth toward networks containing millions or hundreds of millions of nodes and vertices.
  • Information Fusion & Integration: Most tools use complex graph representations and process a medium range of data types, motivating further research on information fusion across formats and, when possible, OSNs.The Information Fusion & Integration dimension measures this capability through the dV ariety(t) metric.
  • Visualization: Visualization capabilities are more evenly distributed than other dimensions, yet few tools combine high Scalability and Visualization capabilities, leaving visualization an open problem.The paper identifies dynamic community finding, data analytics, and pattern finding as future improvement areas.
  • Limitations: The proposed dimensions and degrees are an initial, incomplete framework that may omit relevant characteristics and should be interpreted relative to the current technology and evaluation rubric.The authors expect the values to change as SNA technologies develop.

Appendix: Open Access to Social Network Analysis Dimensions

The paper provides an open-access website and repository to support access to its materials and encourage collaborative evaluation of SNA technologies. These resources are intended to facilitate community participation and future development.

  • Access and collaboration: The website gives researchers, analysts, developers, and engineers access to the data used in the article and supports future collaboration.The resource is aimed at participants across social sciences, science, engineering, and professional SNA practice.
  • Repository contents: The repository includes an open GitHub project, evaluation rubric, analyzed-tools data, collaborative README, and the four-dimensions document.These materials support adding software, documents, papers, and technical data related to SNA technology.
  • Community contribution: The authors invite the community to submit evaluations of the tools already studied and of additional SNA technologies.They present cooperation as necessary for promoting SNA use and fostering new developments.
Loading 2002.09485v1…