26
Computational Analytics in SPOD Gennaro Cordasco, Carmine Spagnuolo , Vittorio Scarano

ROUTE-TO-PA Computational Analytics in SPOD

Embed Size (px)

Citation preview

Page 1: ROUTE-TO-PA Computational Analytics in SPOD

Computational Analytics in SPOD Gennaro Cordasco, Carmine Spagnuolo, Vittorio Scarano

Page 2: ROUTE-TO-PA Computational Analytics in SPOD

ROUTE-TO-PA Prato Plenary Meeting

Overview

15/9/2015

IntroductionSome example

Some Metrics and Methods

Social and Data Network Analysis (SDNA) in SPOD

DiscussionGephi: The Open Graph Viz Platform

Network over time

Page 3: ROUTE-TO-PA Computational Analytics in SPOD

ROUTE-TO-PA Prato Plenary Meeting15/9/2015

IntroductionSocial

NetworkAnalysis

Page 4: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

The power of social networks

Page 5: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

What is a social network?

Page 6: ROUTE-TO-PA Computational Analytics in SPOD

•Social network analysis and data mining:– Detection of criminal activity, counter-terrorism, homeland security,

and intelligence– Analysis of relationships within companies– Sociological and anthropological studies– Reciprocal trust schemes such as eBay ratings– Recommended friends on Facebook– Filter or recommend social media content

• – …

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Applications of social networks

Page 7: ROUTE-TO-PA Computational Analytics in SPOD

ROUTE-TO-PA Prato Plenary Meeting15/9/2015

SomeExamples

Page 8: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Friendship network

Page 9: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Email network

Nodes = People Links = Emails Source: orgnet.com

Page 10: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Karate Club

Page 11: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Romantic and Sexual Relations

Page 12: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Football: Spain won 2010’s World Cup

Page 13: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015

What can we learn by using these models?

ROUTE-TO-PA Prato Plenary Meeting

Page 14: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

• Can we mine SPOD data in order to:• support other analysis (like Dialogue

Games)? • encourage users participation?• identify skilled and/or reliable users?• …

• Idea: assuming that we are able to model the data according to some relationships between different entities:

• we can exploit Network analysis tools.

Network analysis in SPOD

Page 15: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Metricsand

Methods

Page 16: ROUTE-TO-PA Computational Analytics in SPOD

• Assumption: important actors are involved with others extensively.

• The problem is find out automatically which are the central actors.

• Metrics:• Closeness Centrality• Betweenness Centrality• Eigenvalue Centrality• Prestige• PageRank Algorithm• Hits Algorithm• …

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Centrality

Page 17: ROUTE-TO-PA Computational Analytics in SPOD

• Methods:o community detection is key to

understanding the structure of networks.

• In SPOD it can be used to• identify users with common interest (and

suggest friendship);• identify related data;• suggest the proper way of visualizing data.

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Community detection

Page 18: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Networkover

theTime

Page 19: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Networks over time• Does the leader of a discussion change over time?

• Is it true that participate actively in a discussion enhances users’ ability?

• Dynamic graph analysis (DGA) is a novel approach for Network analysis.

• DGA analyses how a network and its metrics evolve over time.

Page 20: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

in

Social

SPOD

andData

AnalysisNetwork

Page 21: ROUTE-TO-PA Computational Analytics in SPOD

• It is easy to analyse a snapshot of a Network at time t (Gephi does it for you).• The key point is to identify the Network structure and semantics

• Different goals require dedicated approaches and are based on different networks.• For instance, if we are going to study interaction among users, we need a network

where: • the users are the entities; • the relations are "who communicate with whom”.

• Even in this simple case, we should also define some timing intervals in order to evaluate whether the interactions change along the time.

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Social and Data Network Analysis (SDNA) in SPOD

Page 22: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Social and Data Network Analysis (SDNA) in SPOD

• Another approach could be define a network that map users and data as a Users-Datalets Network.

Page 23: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Gephi

andvisualization

exploration

Page 24: ROUTE-TO-PA Computational Analytics in SPOD

• Gephi is an open-source network analysis and visualization software package.

• Gephi offers the most common metrics for network analysis:

• Betweenness Centrality;• Closeness;• Diameter;• Clustering Coefficient;• PageRank;• Community detection;• …

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Gephi: The Open Graph Viz Platform

Page 25: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Discussion

Page 26: ROUTE-TO-PA Computational Analytics in SPOD

15/9/2015 ROUTE-TO-PA Prato Plenary Meeting

Discussion

• Goals• Networks• Ideas