+ All Categories
Home > Documents > Approachable Network Analysis

Approachable Network Analysis

Date post: 17-Jun-2015
Category:
Upload: jeff-horon
View: 1,014 times
Download: 1 times
Share this document with a friend
Description:
Unlock the power of network structures in your data. Learn how to build and analyze networks to gain insights through relationship analysis. Apply approachable techniques and free, user-friendly software. Transform the data you have into the data you need – from relational databases and unstructured text to common network structures.Jeff detailed his work in the Medical School Grant Review & Analysis Office. Examples will include: Identifying networks of collaborators from eResearch Proposal Management [eRPM PAF] data, discovering networks of concepts in unstructured text, and use cases from other administrative data sets. Jeff’s presentation included:-“Networks 101″ – The basic building blocks of networks-How people in any business unit can apply network analysis-An emphasis on approachable techniques and free, user-friendly software-Strategies for effectively visualizing and sharing network-driven insights-Tools, tips, and tricks
Popular Tags:
46
Approachable Network Analysis Jeff Horon
Transcript
Page 1: Approachable Network Analysis

Approachable Network AnalysisJeff Horon

Page 2: Approachable Network Analysis

Gartner’s Hype Cycle

Source: http://en.wikipedia.org/wiki/File:Gartner_Hype_Cycle.svg

Page 3: Approachable Network Analysis

My Mission – Short Circuit the Hype Cycle

Source: http://en.wikipedia.org/wiki/File:Gartner_Hype_Cycle.svg

Page 4: Approachable Network Analysis

You will leave here with the knowledge skillsYou will leave here with the knowledge, skills, resources, motivation,

and ideas you need to

d t k l ido network analysis todaytoday

with data you probablywith data you probably

already have

Page 5: Approachable Network Analysis

[Social] Network Analysis

So like Facebook? Sort ofSo, like Facebook? Sort of.

B t t k hBut networks are everywhere.

And they aren’t necessarily “social.”

Page 6: Approachable Network Analysis

TopicsTopics

Networks 101Networks 101Your Use CasesT f i Y D tTransforming Your DataFree, User-Friendly SoftwareExamplesQ&AQ&A

Page 7: Approachable Network Analysis

Networks 101Networks 101

Building BlocksBuilding BlocksPutting the Pieces Together – VisualizationM t iMetrics

Page 8: Approachable Network Analysis

Building Blocks

Nodes [Vertices] – People Things IdeasNodes [Vertices] People, Things, Ideas

Links [Edges] – Relationships

or

Page 9: Approachable Network Analysis

Visualization

Page 10: Approachable Network Analysis

Metrics – Degree

HighestHighestDegree

Page 11: Approachable Network Analysis

Metrics – Degree – In-Degree

HighestgIn-Degree“Popular”Popular

Page 12: Approachable Network Analysis

Metrics – Degree – Out-DegreeHighest Out-Degree“Gregarious”g

Page 13: Approachable Network Analysis

Metrics – Betweenness

HighestHighestBetweenness“Bridge”“Commonalities”

Page 14: Approachable Network Analysis

Metrics – Betweenness

Page 15: Approachable Network Analysis

Metrics – Closeness

HighestHighestCloseness“Who could spread a rumor?”

Page 16: Approachable Network Analysis

Metrics – Closeness

Page 17: Approachable Network Analysis

Metrics – Eigenvector Centrality

HighestHighestEigenvector CentralityCentrality“Importance”

Page 18: Approachable Network Analysis

Metrics – Eigenvector Centrality

Page 19: Approachable Network Analysis

RecapD ( di d) N b fDegree (undirected): Number of

connections

In- / Out-Degree (directed): “Popular” / “G i ”“Gregarious”

Betweenness: “Bridges” / “Commonalities”

Closeness: “Rumor starting point”

Eigenvector Centrality: “Importance”

Page 20: Approachable Network Analysis

Your Use Cases – Connect:

People to Other PeoplePeople to Other People

Things/Ideas to Other Things/Ideas

People to Things/Ideas

Page 21: Approachable Network Analysis

If the other attendees are starting to look like this to you…like this to you…

Page 22: Approachable Network Analysis

Transforming Your DataTransforming Your Data

Common Network Data StructuresCommon Network Data StructuresRelational DatabaseU t t d T tUnstructured Text

Page 23: Approachable Network Analysis

Edge List

A list of edges (links)!

A BA CB CB C

Page 24: Approachable Network Analysis

Edge List

A list of edges (links)!

A B A BA C A CB C B CB C B C

Page 25: Approachable Network Analysis

Edge List

A list of edges (links)!

A B A BA C A CB C B CB C B C

AA

B C

Page 26: Approachable Network Analysis

Data You May Already HaveData You May Already Have

Faculty/Staff and Appointing DepartmentsFaculty/Staff and Appointing DepartmentsFaculty/Staff and GroupsP i i l I ti t d S dPrincipal Investigators and Sponsored

ProjectsSponsored Projects and ParticipantsAuthors and Publications

Page 27: Approachable Network Analysis

Adjacency Matrix

A table of each node by each node

A B C DA| x 1 1 0 AB| 1 x 1 0B| 1 x 1 0C| 1 1 x 0 B CD| 0 0 0D| 0 0 0 x

D

Page 28: Approachable Network Analysis

Transforming Relational Database Data

Where your data has unique identifiers and features associated with them such as:features associated with them, such as:

Page 29: Approachable Network Analysis

Transforming Relational Database Data

Join two instances of your table by the unique identifier:unique identifier:

Page 30: Approachable Network Analysis

Transforming Relational Database Data

Query for both instances of the feature, returning:returning:

Page 31: Approachable Network Analysis

Transforming Relational Database DataNetwork analysis software will remove “self-loops”

and duplicate edges:g

Page 32: Approachable Network Analysis

Transforming Relational Database Data

And the resulting visualization might look like:like:

Page 33: Approachable Network Analysis

Unstructured Text

Node: Word or phraseLink: Co-occurrence within a block of textLink: Co-occurrence within a block of text

Suppose we wanted to find co occurrencesSuppose we wanted to find co-occurrences among words in unstructured text and words of interest included “network” andwords of interest included network and “text.”

You can construct a network based upon word co-occurrence in unstructured textword co-occurrence in unstructured text.

Page 34: Approachable Network Analysis

Unstructured TextYou can construct a network based upon

word co-occurrence in unstructured text.

Edge ListEdge List

network texttext network

Page 35: Approachable Network Analysis

Free, User-Friendly Software

NodeXL [http://nodexl.codeplex.com/][ p p ]

-Microsoft Research / University CollaboratorsMicrosoft Research / University Collaborators

-Installs as an Excel 2007 Template-Installs as an Excel 2007 Template

Free easy and powerful with top notch-Free, easy, and powerful with top-notch visualization

Page 36: Approachable Network Analysis

Free, User-Friendly Software

Simple Text/Network Mining p g

-Homegrown Excel/Visual Basic Package-Homegrown Excel/Visual Basic Package

-Tech Transfer [http://techfinder.techtransfer.umich.edu/ -Search for # 4730]

Page 37: Approachable Network Analysis

LiveLiveDemoDemo

Page 38: Approachable Network Analysis

Specific Examples

Things/Ideas and Other Things/Ideas

Concepts and Other Concepts inConcepts and Other Concepts in Publications and Sponsored Project Proposal / Award DataProposal / Award Data

Page 39: Approachable Network Analysis

Concepts and Other Concepts in Publications and Sponsored Project Proposal / Award Data

C tConceptIncreasing Betweenness Centrality

Page 40: Approachable Network Analysis

Specific ExamplesSpecific Examples

People and Things/IdeasPeople and Things/Ideas

People and Sponsored Projects

Authors and Publication ConceptsAuthors and Publication Concepts

Page 41: Approachable Network Analysis

People and Sponsored ProjectsMedical School PI Engineering PIMedical School Project Engineering Project

Page 42: Approachable Network Analysis

Specific ExamplesSpecific Examples

People and Other PeoplePeople and Other People

Co-Participation on Sponsored Projects,Co-Authorship

Page 43: Approachable Network Analysis

Co-Participation on Sponsored Projects, Co-AuthorshipResearcher / Author Active Project + PublicationI i Ei t C t lit A ti P j tIncreasing Eigenvector Centrality Active Project

Publication

Page 44: Approachable Network Analysis

Strategies for CommunicationStrategies for Communication

VisualizationVisualization-Pay attention to node layoutS btl d h d t-Subtly encode as much data as you can

-Include a really simple key

You understand the network dataYou understand the network data, visualization, and metrics + your audience doesn’t = hand deliverdoesn t hand deliver

Page 45: Approachable Network Analysis

Q&A

Page 46: Approachable Network Analysis

ResourcesResourceshttp://nodexl.codeplex.com/ p p

http://www.umich.edu/~jhoron

Tech Transfer # 4730

On Campus: School of Information, Center for Positive Organizational ScholarshipPositive Organizational Scholarship, Interdisciplinary Group for Research on Innovation


Recommended