Structure, Tie Persistence and Event Detection

1
Structure, Tie Persistence and
Event Detection in Large Phone and
SMS Networks
Leman Akoglu and Bhavana Dalvi lakoglu , bbd _at_cs.cmu.edu Carnegie Mellon University and iLab
Event Detection
Phone and SMS network
Tie Persistence
Around 2M users, 50M edges, 500M phone
calls/SMS 6 months data

• Methodology
• Feature extraction
• Characterize nodes with 12 network-features F
  degree (number of contacts), total weight (phone
  call duration),
• One TxN time-series matrix per feature, T183
  days N1,8M users

• Which link and node attributes are important in
  prediction?

Tie Attributes Node Attributes
Reciprocity (R) 1 if the tie is reciprocal in time tick Degree (K)
Topological Overlap (TO) Cluster Coefficient (C)
Topological Overlap (TO) User reciprocity (r) Faction of ties containing both incoming and outgoing calls
triads in which node is involved
common neighbours
1
2
Node degree
3
4
(left) Z score vs time with W5 and Finweight
(number of calls received). Top 10 days with the
largest Z score is highlighted in red bars.
(middle) u(t) vs r(t-1) for each node at TDec
26th. Top 5 nodes with the largest change is
marked with red stars. (right) inweight vs time
for the top 5 nodes marked notice the change in
calling behavior during the Christmas week.
Tie strength based on (a) SMS
(b) Phone calls
(c) Duration of phone calls


Dataset used for this work was provided by
iLab at Carnegie Mellon University.