Evaluating the Regularity of Human Behavior from Mobile Phone Usage Logs

This paper investigated the relationship between incrementally logged phone logs and self-reported survey data to derive regularity and predictability from mobile phone usage logs. First, we extracted information not from a single value such as location or call logs, but from multivariate contextual logs. Then we considered the changing pattern of the incrementally logged information over time. To evaluate the patterns of human behavior, we applied entropy changes and the duplicated instances ratios from the stream of mobile phone usage logs. By applying the Hidden Markov Models to the patterns, the accumulated log patterns were classified according to the self-reported survey data. This research confirmed that regularity and predictability of human behavior can be evaluated by mobile phone usages.