Adequacy of Data for Mining Individual Friendship Pattern from Cellular Phone Call Logs

In this paper, we study a new problem of mining individual friendship pattern (IFP) for characterizing the interaction behavior of each user in cellular phone call logs. The IFP represents the user's recent frequent relationships and their importance in social lives, which is a unique feature like fingerprint and is useful for many applications such as user resolution and viral marketing etc. We first give the definition and the efficient mining algorithm of the IFP. Then we solve the problem that how much data or time is adequate for characterizing the pattern by introducing a concept of the stable time and a hybrid similarity measure between the IFPs. The experimental result on the real massive cellular phone call logs demonstrates that most users' IFPs can be characterized by a small data set or time.