A Clustering Method Based on Time Heat Map in Mobile Social Network

摘要

Mobile Social Network Services (MSNS) have collected massive amount of users' daily positioning information, which could be used for data mining to learn people's habits and behaviors. This paper proposes a novel clustering method to group nodes according to timestamp information, by analyzing the Time Heat Map (THM), i.e. the activity level distribution of a node during different time intervals. We have employed large amounts of anonymized positioning records coming from a real MSNS, which has extinguished this paper from other researches that use volunteers' daily GPS data. Experiment results have shown that this method not only reveals some interesting features of human activities in real world, but also can reflect clusters' geographical "interest fingerprints" affectively.