Temporal Predictability of Online Behavior in Foursquare

With the widespread use of Internet technologies, online behaviors play a more and more important role in humans’ daily lives. Knowing the times when humans perform their next online activities can be quite valuable for developing better online services, which prompts us to wonder whether the times of users’ next online activities are predictable. In this paper, we investigate the temporal predictability in human online activities through exploiting the dataset from the social network Foursquare. Through discretizing the inter-event times of users’ Foursquare activities into symbols, we map each user’s inter-event time sequence to a sequence of inter-event time symbols. By applying the information-theoretic method to the sequences of inter-event time symbols, we show that for a user’s Foursquare activities, knowing the time interval between the current activity and the previous activity decreases the entropy of the time interval between the next activity and current activity, i.e., the time of the user’s next Foursquare activity is predictable. Much of the predictability is explained by the equal-interval repeat; that is, users perform consecutive Foursquare activities with approximately equal time intervals. On the other hand, the unequal-interval preference, i.e., the preference of performing Foursquare activities with a fixed time interval after another given time interval, is also an origin for predictability. Furthermore, our results reveal that the Foursquare activities on weekdays have a higher temporal predictability than those on weekends and that users’ Foursquare activity is more temporally predictable if his/her previous activity is performed in a location that he/she visits more frequently