Utilization of Graph Analysis in Research and Development (R&D) Tasks
Hello, We are CYRAM Inc., a leading company in Graph Data Science technology.
Utilizing graph analysis in your R&D tasks can enhance both efficiency and effectiveness. Therefore, we would like to explain how graph analysis can be utilized in R&D tasks.
You can also find a summarized version in table format in the attached materials of this post, so we appreciate your interest :)
We hope the materials we provide will be of great help to your work.
[Utilization of Graph Analysis in Research and Development (R&D) Tasks]
Research and Development (R&D) tasks are generally composed of the following key stages.
Project Planning (Research Design)
Research Execution
Results Sharing/Evaluation
Post-Management
At each stage, utilizing Social Network Analysis (SNA), Graph Machine Learning, and Text Network Analysis can enhance efficiency and effectiveness. Below, we explain the analysis topics, required data, and methods for each stage using graph analysis techniques.
※ The Social Network Analysis and Text Network Analysis features are supported in NetMiner4, while most Graph Machine Learning functionalities are available in NetMiner365.
1. Project Planning (Research Design)
In the project planning stage of R&D tasks, graph analysis methods can be utilized to select research topics and set strategic research directions.
Social Network Analysis(SNA)
Analysis Topic: Identifying key researchers and research groups through collaboration network analysis.
Required Data: Co-authorship relationship data and collaboration records among researchers.
Analysis Method:
Model the co-authorship relationships between researchers as a network.
Utilize network centrality metrics (e.g., degree centrality, betweenness centrality) and cohesion group analyses (e.g., community detection) to identify influential researchers and research groups.
This enables the identification of potential collaboration partners and the formation of optimized research teams.
Text Network Analysis
Analysis Topic: Understanding research trends through keyword network analysis of academic literature.
Required Data: Text data from academic papers, patent documents, and technical reports in the relevant field.
Analysis Method:
Extract key terms and topics from the documents using natural language processing (NLP) techniques.
Construct a network based on the co-occurrence relationships between keywords.
Utilize network analysis and topic modeling to identify relationships and clusters among research topics, determine the prominence of key themes in the current field, and uncover structural gaps that highlight new research opportunities.
Graph Machine Learning Analysis
Analysis Topic: Predicting Promising Research Topics Using Citation Networks
Required Data: Citation relationship data of academic papers
Analysis Method:
Model the citation relationships among papers as a graph.
Apply machine learning techniques, such as Graph Neural Networks (GNN), to embed nodes (papers) into a low-dimensional space.
Utilize link prediction methods to identify papers likely to be cited in the future and predict promising research topics based on these insights.
Use Case
Network Analysis-Based Study of ICT Policy Trends
Client: Public Institution in the ICT R&D Sector
Key Objective: Analyze strategic reports on national Information and Communication Technology (ICT) research and development (R&D) to identify changes and differences in the direction and strategies of ICT R&D policies between 2013 and 2018
Key Details
Data : ICT R&D strategic reports from the years 2013 and 2018
Data Preprocessing : Exclusion of irrelevant data, handling thesaurus/defined words, morphological analysis
Data Analysis
Word network creation: measuring co-occurrence frequency between words
Network characteristics analysis: identifying differences between the two periods
Network map (visualization)
Keyword analysis: applying degree centrality and betweenness centrality
Topic modeling: identifying themes within documents
Utilization of results
Comparison of ICT R&D directions and changes in detailed implementation plans between reports written in two different periods.
Identification of trends and trend changes between the two periods.
[Word cloud for the 2013 and 2018 strategic reports]
2. Research Execution
In the research execution stage of R&D tasks, graph analysis methods can be used to enhance team collaboration efficiency, effectively monitor research progress, and improve the accuracy of experimental results.
Social Network Analysis(SNA)
Analysis Topic: Enhancing communication efficiency through the analysis of collaboration structures within the research team.
Required Data: Emails, meeting notes, collaboration tool usage records, or surveys among research team members.
Analysis Method:
Collect communication data among research team members to construct a network.
Analyze centrality metrics (degree centrality, closeness centrality, betweenness centrality, etc.) to identify key individuals in the information flow (hubs, intermediaries, etc.) and potential problem areas.
This enables the identification of communication bottlenecks and the improvement of collaboration structures, thereby enhancing the communication efficiency and productivity of the research team.
Text Network Analysis
Analysis Topic: Monitoring research progress through network analysis of key concepts in research notes and experiment records.
Required Data: Text data such as research notes, experiment reports, and interim presentation materials.
Analysis Method:
Extract key concepts and keywords from documents generated during the research process using natural language processing techniques.
Construct a network based on the co-occurrence relationships between keywords.
Visualize the progress of each research stage through network analysis and utilize the insights to adjust the research's substantive direction.
Graph Machine Learning Analysis
Analysis Topic: Predicting outcomes through modeling complex interactions within experimental data.
Required Data: Relationship data between experimental variables and outcomes.
Analysis Method:
Construct a graph with experimental variables and outcomes as nodes and the relationships between variables as links.
Apply machine learning techniques such as Graph Neural Networks (GNN) to embed nodes into a low-dimensional space.
Utilize the trained model to predict outcomes under new experimental conditions and optimize experimental design.
Use Case
Social Network Analysis (SNA)-based organizational collaboration and communication network analysis.
Client: Automobile manufacturing company
Key Objective:
Identifying the current state of collaboration and communication networks within the organization, including information sharing, cooperation, and command/reporting
Identifying and addressing collaboration issues within the organization
Key Details
Visualizing communication networks by department and understanding communication structures through network index measurements
Identifying influencers and marginalized individuals through communication impact analysis
Utilization/Expected Benefits
Gaining insights for understanding and adjusting collaboration and communication structures
Enabling cross-organizational comparisons through quantitative measurements
Collaboration and communication management through regular measurements
[Visualization of organizational collaboration and communication networks (example)]
3. Results Sharing/Evaluation
In the performance sharing and evaluation stage of R&D tasks, graph analysis can be utilized to identify the dissemination pathways and impact of research outcomes, effectively convey key insights, and strategically set future research directions through objective performance evaluation.
Social Network Analysis(SNA)
Analysis Topic: Impact evaluation through analysis of research outcome dissemination pathways
Required Data: Citation data on research results, social media mention data, and collaboration network data
Analysis Method:
Collect data on when and how research outcomes are cited and mentioned in academic papers, patents, and social media.
Construct a network based on these citation and mention relationships, and analyze connectivity and centrality metrics (e.g., network diameter, path length, eigenvector centrality) to evaluate the dissemination pathways and impact of research outcomes.
This enables understanding the ripple effects of research outcomes (e.g., scale, speed) and utilizing these insights to set future research directions.
Text Network Analysis
Analysis Topic: Deriving key content (themes) through network analysis of major concepts in research outcome documents
Required Data: Text data from research reports, papers, and presentation materials
Analysis Method:
Extract key keywords from research outcome documents using natural language processing techniques.
Construct a network based on the co-occurrence relationships between keywords.
Identify the key thematic keywords of research outcomes through network analysis (e.g., centrality analysis, cohesion group analysis) and develop effective communication strategies.
Graph Machine Learning Analysis
Analysis Topic: Developing a performance prediction model through analysis of relationships among research outcome evaluation metrics
Required Data: Research outcome evaluation metric data (e.g., number of papers, citations, patents, technology transfer achievements, etc.)
Analysis Method:
Model the relationships among research outcome evaluation metrics as a graph.
Apply machine learning techniques, such as Graph Convolutional Networks (GCN), to embed nodes (evaluation metrics).
Utilize the trained model to predict how changes in specific evaluation metrics affect others and apply these insights to assess overall research performance comprehensively.
Use Case
Planning and development of a foundational and basic research outcome information provision system
Client: National R&D project information management agency
Key Objective:
Establish a knowledge map* to allow entrepreneurs (general public), researchers, and policymakers to easily and effectively/efficiently search and utilize key performance information derived from basic and foundational research projects.(A knowledge map visually represents the relationships between knowledge and information as a network by effectively analyzing and processing useful information that is hidden or latent within the data.)
Key Details
Data
Project information
Implementing organization information
Implementing personnel information
Outcome information (Outcome report)
Knowledge map modeling
Expected benefits
By identifying the connection structure between outcomes derived from search results, it helps determine priorities for further searches, effectively supporting knowledge exploration
Providing suitable information and indicators for science and technology research planning, evaluation, investment, and policy development
[Knowledge map]
4. Post-Management
In the post-management stage of R&D tasks, continuous utilization of research outcomes and suggestions for improvements are provided, maximizing the impact of the research.
Social Network Analysis(SNA)
Analysis Topic: Dissemination of research outcomes and collaboration network analysis
Required Data: Citation data of research outcomes, collaboration data between researchers, conference and seminar attendance records, etc.
Analysis Method:
Collect data on how research outcomes are cited and utilized in academia and industry.
Construct a collaboration network among researchers and analyze centrality, cohesion, and role group metrics.
This helps identify the impact of research outcomes and collaboration structures, and is used to develop future collaboration strategies.
Text Network Analysis
Analysis Topic: Analysis of feedback and reviews related to research outcomes
Required Data: Text data from feedback documents on research outcomes, review papers, media articles, etc.
Analysis Method:
Extract key keywords and topics from feedback and review documents.
Construct a network based on the co-occurrence relationships between keywords.
Through network analysis, identify key opinions and areas for improvement regarding research outcomes, and use these insights to set future research directions.
Graph Machine Learning Analysis
Analysis Topic: Predicting the commercialization and technology transfer potential of research outcomes
Required Data: Patent data related to research outcomes, market trend data, technology transfer case data, etc.
Analysis Method:
Model patent and market data related to research outcomes as a graph.
Apply machine learning techniques, such as GraphSAGE, to embed nodes (patents, technologies) into a low-dimensional space.
Utilize the trained model to predict the commercialization potential and technology transfer feasibility of research outcomes, and apply these insights to strategic decision-making.
Use Case
Performance information analysis using Social Network Analysis (SNA)
Client: Central administrative agency
Key Objective:
Introduction of Social Network Analysis (SNA) techniques to analyze R&D performance information in response to R&D project evaluations
Establishing strategies for the dissemination of future performance information and building a service system
Effective utilization of performance information by policymakers and researchers
Key Details
Data
Project information
Performance information (papers, intellectual property rights, royalties, commercialization, etc.)
Data Modeling
Utilization of results
Visualize hidden relationships between researchers and research fields formed through R&D projects and outcomes, making it easier for future researchers to collaborate with potential co-researchers and identify trends between research fields
This allows policymakers to identify trends between research fields and analyze the current state of technological linkages between them, while researchers can identify key individuals among their peers, understand the ripple effects of research outcomes, and use these insights to set future research directions
[Network map between research fields]