Pervasive Stress Recognition for Sustainable Living
Author(s)
Bogomolov, Andrey; Lepri, Bruno; Ferron, Michela; Pianesi, Fabio; Pentland, Alexander Sandy
DownloadAccepted version (200.9Kb)
Terms of use
Metadata
Show full item recordAbstract
In this paper we provide the evidence that daily stress can be reliably recognized based on human behavior metrics derived from the mobile phone activity (call log, sms log, bluetooth interactions). We introduce an original approach for feature extraction, selection, recognition model training and discuss the experimental results based on Random Forest and Gradient Boosted Machine algorithms. Random Forest based model showed low variance comparing to the GBM-based one, thus winning the bias-variance tradeoff and preventing over-fitting, given the noisy source data. Potential impact of the technology is reducing stress and enhancing subjective well-being for sustainable living. © 2014 IEEE.
Date issued
2014-03Department
Massachusetts Institute of Technology. Media LaboratoryPublisher
Institute of Electrical and Electronics Engineers (IEEE)
Citation
Bogomolov, Andrey, Lepri, Bruno, Ferron, Michela, Pianesi, Fabio and Pentland, Alex Sandy. 2014. "Pervasive Stress Recognition for Sustainable Living."
Version: Author's final manuscript