Time division based on analyses of network user time span preference

The time based feature facilitates the new opportunity to understand the network user behaviors. Research on user behavior patterns varying with time is a hot point recently. In multi-dimensional mining of user profiles, time division contributes to decline the complexity of data process. Time division consistent with user on-line regularities is significant to obtain accurate user behavior information. The time characteristics of Internet applications appear as users´ preference to time spans when it comes to user behaviors. In this paper we introduce hour-based time division method considering the user time span preference. First, we arrange users into 24 groups, in which each user has the highest on-line duration in the same 1-hour time span. Then we cluster the 24 groups with the clustering algorithm based on user time span preference. Corresponding to the result groups, we achieve the time division by incorporating 1-hour time spans. Finally, a set of useful descriptions and analyses on user time span preference are presented to give us a deep sight of user behaviors with the result time division.