Graph visualization options and latest developments

Graph Visualizationoptions and latest developments

By @Linkurious - [email protected]

London Neo4j meetup at Skills Matter, Feb 27, 2013

Linkurious startup: a few words about us• Jean Villedieu, CMO Linkurious• Sébastien Heymann, CEO Linkurious• Romain Yon, CTO Linkurious (is coding tonight)

2/17/13 2

Example of graph

2/17/13 3

Graphs are in our DNA• Gephi, the leading graph analysis and exploration solution• Democratizing network thinking• Linkurious

2/17/13 4

The best way to store and access the data of a connected world

• Graphs are “everywhere”: conceptual view on data• Neo4j and graph databases offer a fast and intuitive way to use data• Social networks, recommendation engines, networked systems…

2/17/13 5

Visualization is key to understanding graphs• Verify data• Check algorithm• View information in context• Build a visual representation• Understand relationships• Discover information• Find patterns• Illustrate and communicate data

2/17/13 6

Global approach vs Centred approachCard, Shneiderman (Using Vision to Think, 1999)

1. Overview first,2. Zoom and filter,3. Details-on-demand.

7

Van Ham, Perer (2009)

1. Search, 2. Show Context, 3. Expand on Demand.

Play now: which tool for which approach?

Cytoscape, D3.js, Gephi, GraphViz, KeyLines, Linkurious, MentionMap,

Neo4j server web interface, Neoclipse, NodeXL, Palantir,

Sigma.js, SocialAction, Wikisway, WikiWeb...

8

WikiWeb

D3.js

Toolkits

Summary: which tool for which approach?Cytoscape, GraphViz,

Gephi, GraphInsight, GUESS,NodeXL, SocialAction, Tulip,

UCINET

Global approach

18

Linkurious, MentionMap, Neo4j server web interface,

Neoclipse, Wikisway, WikiWeb

Centred approach

+ Toolkits to build your own App: D3.js, KeyLines, Sigma.js

Global approach vs Centred approachGreat for:

Identifying communities and global properties

Finding correlations properties / structure

Communicating « wow » effect

19

Great for:

Investigating entities

Exploring local context

Communicating insights

Global approach vs Centred approachIssues:

Difficult to learn (data mining + CS + design skills)

Poor insights with large graphs

Highly sensitive to graph incompleteness (<5% missing data => important bias)

Hardware limits (goodbye mobile!)

20

Issues:

Users must know what to search

Poor statistical insights (bias due to hidden information)

Orientation is difficult

Global approach vs Centred approachUse it for:

Analysis at the global scale.

Make hypotheses or try to confirm statistical properties.

Graphs of size < 50k nodes.

21

Use it for:

Database browsing to quickly find information.

Analysis of local views.

Graphs of any size.

3 applications of the centred approach

Use case 1: fraud detectionAutomatic outlier detection over time (Seb's Ph.D.)

Visualization of outlier-centred graphs

23

Use case 2: visual recommendationYasiv.com

24

Use case 3: (telco) impact analysisNetwork: millions of routers and machinesLocal impact of a router failure in network services?

25

Entypo pictograms by Daniel Bruce — www.entypo.com

Linkurious: visualize graph data easily• A tool to visualize any graph database• Graph search made easy• Approach : search, show context, expand on demand

2/17/13 27

Linkurious competitorsLocked to a specific use case:MentionMap => Twitter mentionsPalantir => for Palantir customers onlyTheAdvisor => scientific papers recommendationWikiways => TechCrunch DatabaseWikiweb => Wikipedia browsing

Limited browsing features:Neoclipse, Neo4j web admin

Toolkits (require to code): D3.js, KeyLines, Sigma.js

28

Live demo

2/17/13 29

Questions?

@[email protected]