Mining the Connected World

Mining the Connected WorldEe-Peng LIM

Director, Living Analytics Research CentreProfessor, School of Information Systemsy

http://larc.smu.edu.sg

Fraunhofer IDM@NTU Workshop, 20 February 2012

Simple Statistics• How many of us are on Facebook today?

> 845 million (December 31, 2011)

• How many of us are on Twitter today?

> 300 million (June 2011)( )

Living Analytics Research Centre

Living Analytics =

Consumer & Social Insights From

Experiment-Driven Closed-Loop Analytics +g ySocietal Scale Human Networks

LARC D t S ttiLARC Data Settings

A Glimpse of LARC Research:

(a) Mining Link Formation Rules

Link Formation Rule Mining:Do relationships lead to other relationships?Do relationships lead to other relationships? • Local structures for understanding and predicting the

dynamics of large complex networks

All possible triads in a directed graph

• Previous research however does not consider the formation order of links

• We therefore study local structures for link formation in directed, labeled, temporal social networks

Link Formation Rules (LF-Rules)

• LF-rule: Rule of a node (user) forming new links to other nodes (users) based on pre-existing local link structures.

precondition The link from s to e is formed precondition as a postcondition

Mining Methodology

• Mine LF-rules from a social network with temporal links• Mine LF-rules from a social network with temporal links.• Apply randomizing technique to the network, for

estimating the expected support of LF-rules in a random graph

• Evaluate interesting rules with higher-than-expected supportsupport

Interesting LF-rules in myGamma

• Based on the Dec 2009 snapshot690k ith t l t 1 li k– ~690k users with at least 1 link

– > 9 million links (~93% friend links)

• Top 5-rules in terms of support

Interestingness scoressupport expected

supportsurprise

(supp/exp. supp)confidence

28.91% 22.41% 1.29 43.22%

28.38% 22.37% 1.27 43.1%

25 42% 13 54% 1 88 39 15%25.42% 13.54% 1.88 39.15%

24 37% 1 22% 20 06 31 98%24.37% 1.22% 20.06 31.98%

20.55% 11.49% 1.79 27.52%20.55% 11.49% 1.79 27.52%

Major Observations• Users tend to rely more on mutually trusted

friends in forming new friendship links. – R12 (right) has much higher confidence (~34% vs.

~22%) and surprise values (5.32 vs. 3.52) than R11(left)(left)

• 3.45% of users reciprocated a friend link with a pfoe link.

A Glimpse of LARC Research:A Glimpse of LARC Research:

(b) Palanteer: A Data Analytics Engine for Twitter DataEngine for Twitter Data

Palanteerhttp://palanteer.sis.smu.edu.sghttp://palanteer.sis.smu.edu.sg

tranportation

Search Box

Trending items

E t J l 12 2011Event on July 12, 2011

MRT Event

How do Singapore users feel?

How popular is Starbucks?

Palanteer – Taiwan Edition

Palanteer – Thai Edition

Conclusions• Interesting research problems in the

connected worldco ected o d• Living analytics focuses on discovering

user preferences friendship patterns anduser preferences, friendship patterns, and trends

• Living analytics is multidisciplinary• Living analytics is multidisciplinary• LARC looks forward to exciting

ll b ti ith i d t t dcollaborations with industry partners and other researchers

LARC Activities

Thank youEe-Peng LIM

http://larc.smu.edu.sg

AcknowledgmentFaculty Members: Jing JIANG, Feida ZHU, David LO, Hady LAUW

Collaborators (NTU) : Aixin SUN, Marko SKORIC, Anwitaman DATTAResearchers: Cane LEUNG, Aek Palakorn, Bingtian DAI, Agus, Nelman

PhD Students: Freddy, Hanbo, Tuan Anh, Minh Duc