Modeling physical activity propagation, such as the activity level and intensity, is the key to prevent the cascades of obesity, and help spread wellness and healthy behavior in a social network. However, there has been lacking of scientific and quantitative study to elucidate how social communication may deliver physical activity interventions. In this work we introduce a Community-level Physical Activity Propagation (CPP) model to analyze physical activity propagation and social influence at different granularities (i.e., individual level and community level). CPP is a novel model which is inspired by the well-known Independent Cascade and Community-level Social Influence models. Given a social network, we utilize a hierarchical approach to detect a set of communities and their reciprocal influence strength of physical activities. CPP provides a powerful tool to discover, summarize, and investigate influence patterns of physical activities in a health social network. The detail experimental evaluation shows not only the effectiveness of our approach but also the correlation of the detected communities with various health outcome measures (i.e., both existing ones and our novel measure, named Wellness score, which is a combination of lifestyle parameters, biometrics, and biomarkers). Our promising results potentially pave a way for knowledge discovery in health social networks. NhatHai Phan CIS Department, University of Oregon, Eugene, OR

1. ACM International Conference on Information and Knowledge Management (CIKM) - 2014
Analysis of Physical Activity Propagation
in a Health Social Network
Nhathai Phan, Dejing Dou, Xiao Xiao,
Brigitte Piniewski, David Kil
1

2. Outline
• SMASH Project & Motivation
• Community-Level Physical Activity Propagation
• Experimental Results
• Conclusions & Future Works
2

3. Obesity & Physical Activity Interventions
• 18 states (30% <35%), 2 states (>= 35%)
• Medical cost:
– $147 billion
(in 2008)
• 30 minutes, 5 days
• Interventions
– Telephone (16)
– Website (15)
– Effective in
short term
3
Prevalence* of Self-Reported Obesity Among U.S. Adults
CDC, http://www.cdc.gov/obesity/data/prevalence-maps.html-2014
E.G. Eakin et al. 2007
C. Vadelanotte et al. 2007
G.J. Norman et al. 2007

4. SMASH Project
• 254 Overweight and Obese individuals with personal
information in the YesiWell study
• Social activities
– Online social network, text messages, posts, comments, …
– Social games, competitions, …
• Daily physical activities
– Walking, running, jogging, distance, speed, intensity, …
• Biomarkers, biometric measures
– Cholesterol, triglyceride, BMI, …
4

5. Motivation
• Utilize social networks to
• help the physical activity propagation process
• improve the intervention approaches with
affordable cost
• How can social communications effect the
physical activity propagations?
– Social interactions
– Different granularities
– Physical activity propagations & health outcomes
5

6. Outline
• SMASH Project & Motivation
• Community-Level Physical Activity Propagation
• Experimental Results
• Conclusions & Future Works
6

7. ])([#])([# ttstepsttsteps 
…………………….…………………….
A Trace of Physical Activity
Propagation
7
m, t
v
u
[t, t+tw]

8. Problem Statement
• A directed graph
– represents an influence relationship
– represents the strength of the arc
• A set of traces
8K. Saito, R. Nakano, and M. Kimura. Prediction of information diffusion probabilities for independent cascade model. In KES’08, pages 67-75.
Y. Mehmood, N. Barbieri, F. Bonchi, and A. Ukkonen. Csi: Community-level social inuence analysis. In ECML-PKDD’13, pages 48-63.
CPP Model

9. CPP Model Definition (1)
• Log likelihood of the traces given
• Users’ responsibility:
9
p

10. CPP Model Definition (2)
• CPP model learning
• Probability function
• is a selection function
10
f

11. Learning & Model Selection (1)
• Complete expectation log likelihood of the
observed propagations:
• Solving
• We have
11

12. Learning & Model Selection (2)
• Users’ responsibilities
will not change
• Run EM algorithm
without clustering
structure
– step 1: estimate
– step 2: update
• Keep fixed,
update
• Bayesian Information
Criterion (BIC)
12

13. Outline
• SMASH Project & Motivation
• Community-Level Physical Activity Propagation
• Experimental Results
• Conclusions & Future Works
13

14. Experiment Setting
• YesiWell dataset
– 254 users
– Oct 2010 – Aug 2011
• BMI value
• Wellness score
• Parameter setting:
– tw is a day, is a week
14
2
))((
)(
mheight
kgmass
BMI 
)1()(
)/()(
3423
121
cHbAULDLU
HDLTGUBMIUy




t

15. Detected Communities
• Influencers: circle nodes
• Influenced users: rectangle nodes
• Non-Influenced users: triangle nodes
15

16. Detected Communities
with Health Outcome Measures
16
avg(BMI) avg(WS)
avg(#steps)

17. Consistency of Detected Communities
17
Standard deviation of BMI Standard deviation of WS

18. CPP vs Social Link, CSI Model
• Apply optimal clustering on friend network
18
Wellness score #steps

19. Outline
• SMASH Project & Motivation
• Community-Level Physical Activity Propagation
• Experimental Results
• Conclusions & Future Works
19

20. Conclusions and Future Works
• Propose the CPP model
• Observations:
– Social networks have great potential to propagate
physical activities
– The propagation network found is almost acyclic
– The physical activity-based influence behavior has a
strong correlation to health outcome measures (BMI,
lifestyles, and Wellness score)
• Which types of messages are important?
• Which messages could influence non-influenced
users?
20
 






Vu
Z
w
Iv
Z
w
m
muvm
u
m
muvm
uvpuvppL
,,,
,
,,,
1
)),(1(),()(


 


21. ACM International Conference on Information and Knowledge Management (CIKM) - 2014
Thanks you!
{haiphan, dou}@cs.uoregon.edu
21