Topic-Aware Physical Activity Propagation in a Health Social Network

Phan, NhatHai; Ebrahimi, Javid; Kil, Dave; Piniewski, Brigitte; Dou, Dejing

doi:10.1109/mis.2015.92

Cited by 18 publications

(10 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…whereb t k andĉ t are dynamic biases, B k e is a matrix of weights which connects x t u with hidden variables in the group k. ψ t (v, u) is the probability that v influences u on physical activity at time t. ψ t (v, u) is derived from the TaCPP model [60].…”

Section: Human Behavior Modelingmentioning

confidence: 99%

Preserving differential privacy in convolutional deep belief networks

Phan

Dou

2017

Mach Learn

Self Cite

View full text Add to dashboard Cite

The remarkable development of deep learning in medicine and healthcare domain presents obvious privacy issues, when deep neural networks are built on users’ personal and highly sensitive data, e.g., clinical records, user profiles, biomedical images, etc. However, only a few scientific studies on preserving privacy in deep learning have been conducted. In this paper, we focus on developing a private convolutional deep belief network (pCDBN), which essentially is a convolutional deep belief network (CDBN) under differential privacy. Our main idea of enforcing ϵ-differential privacy is to leverage the functional mechanism to perturb the energy-based objective functions of traditional CDBNs, rather than their results. One key contribution of this work is that we propose the use of Chebyshev expansion to derive the approximate polynomial representation of objective functions. Our theoretical analysis shows that we can further derive the sensitivity and error bounds of the approximate polynomial representation. As a result, preserving differential privacy in CDBNs is feasible. We applied our model in a health social network, i.e., YesiWell data, and in a handwriting digit dataset, i.e., MNIST data, for human behavior prediction, human behavior classification, and handwriting digit recognition tasks. Theoretical analysis and rigorous experimental evaluations show that the pCDBN is highly effective. It significantly outperforms existing solutions.

show abstract

Section: Human Behavior Modelingmentioning

confidence: 99%

Preserving differential privacy in convolutional deep belief networks

Phan

Dou

2017

Mach Learn

Self Cite

View full text Add to dashboard Cite

show abstract

“…Their findings show that: (a) social networks have great potential to propagate physical activities via social communications; and (b) physical activity-based influence behavior has a strong correlation to health outcome measures such as BMI. We later showed that the influence of a post is related with the distribution of the topics in the content of the post [26]. Moreover, in [24] a deep learning model for human behavior prediction in health social networks was proposed.…”

Section: Physical Activity Interventionmentioning

confidence: 99%

“…To model physical activity propagation, Phan et al [25] introduced the Community-level Physical Activity Propagation (CPP) model, which was inspired by the ideas of the Independent Cascade (IC) model [17], and the Community-level Social Influence (CSI) model [21]. Similar to our Topic aware Community-level Physical Activity Propagation (TaCPP) model [26], we take the content of the posts into account. However, to study the physical activity of the network better, the current work is based on the labels of activitydisclosure posts and not the topics of all the posts available.…”

Section: Group-level Behaviormentioning

confidence: 99%

See 1 more Smart Citation

Characterizing Physical Activity in a Health Social Network

Ebrahimi

Phan

Dou

et al. 2016

Proceedings of the 6th International Conference on Digital Health Conference

Self Cite

View full text Add to dashboard Cite

New horizons are emerging within healthcare delivery, education, intervention provision, and tracking. We study a health social network that has tracked physical activities, biomarkers, and posts the participants have shared, throughout a one-year program. The program was aimed at helping people to adopt healthy behaviors and to lose weight. In this paper, we focus on users' posts that relate to physical activities. Prior papers characterize health based solely on users' information disclosed through natural language or questionnaires. The drawback of these works is their lack of medical records or health-related information to validate their findings. By contrast, with our direct access to users' physical and medical data, we investigate the implication of users' posts at both individual and group levels. We are able to validate our hypotheses about the effects of certain social network activities, by contextualizing them in the specific users' actual medical progress and documented levels of exercise. Our findings show that activity self-disclosure posts are good indicators of one's realworld physical activity, which makes them good resources for monitoring the participants. In addition, using a physical activity propagation model, we show how these posts can influence the physical activity behavior at the network level. Further, posts exhibit distinctive affective, biological, and linguistic style markers. We observe that these characteristics can be used in a predictive capacity, to detect positive activity signals with ∼ 88% accuracy, which can be utilized for an unobtrusive monitoring solution.

show abstract

“…Gojaznost je štetna po zdravlje jer vodi ka bolestima kao sto su: dijabetes, bolest srca i povišen krvni pritisak [4]. Sa druge strane, gojaznost vodi i ka mentalnim bolestima kao što su: usamljenost, anksioznost i depresija.…”

unclassified

Using probability density function in the procedure for recognition of the type of physical exercise

Cakić¹,

Dilparic²,

Zigic³

et al. 2017

Zb EI Nikola Tesla

View full text Add to dashboard Cite

Kratak sadržaj: U radu je prikazan postupak za prepoznavanje fizičkih vežbi, korišćenjem samo troosnog akcelerometra pametnog telefona. Telefon se nalazi nefiksiran u džepu vežbača. Analizirane su vežbe za jačanje mišića nogu iz stojećeg položaja: čučanj, iskorak i podizanje kolena. Sve vežbe su rađene nogom u čijem džepu se nalazi senzor ubrzanja korišćenog mobilnog telefona. Da bi se testirao predloženi postupak, nasumice je odabrana vežba podizanje kolena koja se obavlja suprotnom nogom od one u čijem se džepu nalazi mobilni telefon. Filtriranje sirovih signala ubrzanja je postignutno korišćenjem Batervortovog filtra propusnika niskih učestanosti (desetog reda). Filtrirani signali svake od osa akcelerometra su opisani pomoću tri deskriptora. Nakon izračunavanja deskriptora signala, za svaki deskriptor je formirana funkcija gustine verovatnoće. Postupak za prepoznavanje vežbi se obavlja online unutar Android aplikacije pametnog telefona. Signali dve muške i dve ženske osobe su poslužili kao referenca za prepoznavanje vežbi. Uspešnost prepoznavanja je 94,22% za prepoznavanje tri vežbe, odnosno 85,33% za prepoznavanje sve četiri razmatrane vežbe.

show abstract

Topic-Aware Physical Activity Propagation in a Health Social Network

Abstract: Modeling physical activity propagation, such as physical exercise level and intensity, is the key to preventing the conduct that can lead to obesity; it can also help spread wellness behavior in a social network.

Cited by 18 publications

References 18 publications

Preserving differential privacy in convolutional deep belief networks

Preserving differential privacy in convolutional deep belief networks

Characterizing Physical Activity in a Health Social Network

Using probability density function in the procedure for recognition of the type of physical exercise

Contact Info

Product

Resources

About