Nathan Eagle, Alex Pentland

Abstract

In this work we identify the structure inherent in daily human behavior with models that can accurately analyze, predict and cluster multimodal data from individuals and groups. We represent this structure by the principal components of the complete behavioral dataset, a set of characteristic vectors we have termed eigenbehaviors. In our model, an individual’s behavior over a specific day can be approximated by a weighted sum of his or her primary eigenbehaviors. When these weights are calculated halfway through a day, they can be used to predict the day’s remaining behaviors with a 79% accuracy for our test subjects. Additionally, we show that users of a similar demographic can be clustered into a “behavior space” spanned by a set of their aggregate eigenbehaviors. These behavior spaces make it possible to determine the behavioral similarity between both individuals and groups, enabling 96% classification accuracy of group affiliations. This approach capitalizes on the large amount of rich data previously captured during the Reality Mining study from mobile phones continuously logging location, proximate people, and communication of 100 subjects at MIT over the course of nine months.