The users' role is crucial in the development, deployment and the success of online social networks (OSNs). Despite this fact, little is known and even less has been published about user activities in the operating OSNs. In this paper, we present a large scale measurement analysis of user behaviour, in terms of time spent online, in some popular OSNs, namely Bebo, Flixster, MySpace, and Skyrock, and characterise user groups in OSNs. We used more than 200 PlanetLab  nodes for our measurement, monitored more than 3000 users for three weeks by downloading repeatedly their profile pages; more than 100 million pages were processed in total. The main findings of the paper are the following. Firstly, we create a measurement framework in order to observe user activity. Secondly, we present cumulative usage statistics of the different OSNs. Thirdly, we classify the monitored users into different groups and characterise the common properties of the members. Finally, we illustrate the wide applicability of our datasets by predicting the sign out method of the OSN users.