Network Visualization and Analysis

Network analysis is a powerful methodology that can be used to study and visualize the relations in data. Using advanced graph theory algorithms, we can identify the clusters of nodes that belong together, reveal the most influential nodes, and discover the structural gaps in a network. This can be used both to get an overview of an interconnected system or to affect it in the most efficient way.

The best dataset for network analysis will consist of relations: transactions, interactions, or co-occurrences. Using InfraNodus, you can create a new graph easily using the live text-to-graph input or import your existing data: CSV spreadsheets or Gephi Gexf files.

You can then obtain the graph analytics for every node as well as the whole network. You can also save and export the graph for further exploration in other platforms, such as Gephi and Pajek, or as an image to add it into your work. Graphs that are made public can also be embedded on other websites or shared via a URL. For instance, the graph below:


This interactive graph above is made using InfraNodus. It contains relations between the organs derived from Traditional Chinese Medicine. You can see the main nodes on the graph (they have high betweenness centrality and are shown bigger on the graph) as well as the main communities (the clusters of nodes that tend to be closely related) and the statistics on every node and on the network in the graph analytics panel (at the bottom right corner). Feel free to play around with this graph to see how it works.


Step-by-Step Workflow for Network Analysis

Here is how you can visualize your own graph or data using InfraNodus. The demonstration below makes use of the live text-to-graph input, which enables you to add the nodes using #hashtags or @mentions.

You can import a graph from Gephi or any other software (Gexf or CSV formats) or simply create your graph yourself. You can also import the data from Twitter, Excel and other sources.

Create a new mind map graph

When you are building your own graph, you can use the #hashtags and @mentions to add the nodes (@ links to all the #). The rest of the text is a description of the relation. For example, the phrase "@liver is responsible for #metabolism and cleans #blood" will create 3 nodes and link the ones marked with a # sign to the one marked with the @ sign.

Use the text-to-graph input to create your network visualization

You will then see the most influential nodes (betweenness centality) and the most prominent communities. All the stats for the nodes and the graph structure are available in the Analytics panel.

See your mind map as a network graph.

Use the Insight feature to get structural insights into the network: what are the main gaps and what nodes and communities could be bridged together.

Structural holes in knowledge graphs generate new ideas


How It Works: the Methodology

There are multiple network analysis tools available today. Some of them, such as Python or R-based libraries like NetworkX, require programming skills. Some other software tools, such as Gephi, are more user-friendly and provide visualization capacities. One of the reasons we developed InfraNodus was to provide a simple-to-use web-based tool where you can use the existing data to understand the network thinking approach. We tried to make it less technical while keeping all capabilities for advanced data analysis.

Network analysis methods implemented in InfraNodus are based on the reputable open-source libraries, such as Cytoscape and Graphology. Our approach is described in detail in our peer-reviewed paper: InfraNodus — Geneating Insight Using Text Network Analysis (Paranyushkin, 2019) presented at The Web Conference in 2019. You can refer to this paper for more detail on the algorithms used or for citation.

Here is a brief outline of the methodology that we recommend:

  1. Choose the dataset that contains relations. It could be a list of interactions in a company (social network), a text (relations between the words), a transportation network (origin and destination), or more complex data.
  2. Import the data into the network analysis tool of your choice
  3. Apply node metrics algorithms to range the nodes by a certain parameter (e.g. InfraNodus uses betweenness centrality or degree — the number of connections a node has — as a measure of influence)
  4. Run a modularity algorithm (Blondel et al based on Louvain) that will identify the communities in the network graph. The nodes (words) that appear more frequently together will belong to the same community and have a distinct color.
  5. Align the words / nodes on a 2D / 3D plane where the most connected nodes (the hubs) are pushed apart from each other, while the less connected nodes are pulled towards hubs (Force-Atlas layout algorithm). This helps build topical clusters that reflect the words' co-occurrence and correlate with the community structure identified in the previous step.
  6. Reflect on the patterns and connectivities obtained (in InfraNodus, you can use the Analytics panel > Essence tab for that).
  7. Look for the structural gaps: which distinct clusters of nodes could be connected but are not connected yet (in InfraNodus, you can use the Analytics > Insight tab for that).
  8. Check the statistics for nodes, look for irregularities. For example, in InfraNodus, we offer the "diversivity" measure, which is betweenness centrality divided by the degree. The higher is the measure for a node, the more influence it has with the least number of connections. In other words, those are "VIP nodes" that do not interact with too many nodes, but the nodes they interact with are very important.
  9. Reveal the "non-obvious" — remove the biggest nodes from the graph, recalculate the metrics, and reveal the new influential nodes and communities.
  10. Reiterate

Try It Yourself

You can try this approach yourself using InfraNodus for any text or graph data:

Log In    Sign Up


Custom Service: Network Visualization

Please, let us know if you have an interesting proposition, use case, research project, or service request.

Contact Us