We further compare the difference between heavy users and normal users in terms of their daily and hourly activity. We can clearly see from Fig. 5(a) that most heavy users (83.84%) activate the data service more than six days in a week. As shown in Fig. 5(b), heavy users generate traffic in 14 hours averagely in a day, 70% and 31% of heavy users use applications for more than 10 and 20 hours respectively.
We can conclude from above analysis that different mobile users tend to have distinct data usage patterns. Top 1% heavy users contribute up to 88% of the total mobile data traffic, and use applications much more frequently.
B. Mobility Pattern
Human mobility pattern is essential for a deep understanding of network dynamics and evolution. User mobility pattern in Mobile Internet also impact the network resource allocation and social network [27]. Here we investigate the users’ mobility patterns when they are active in data service usage.
Due to the cell tower oscillation (a mobile device located at the boundary of two cells may access alternately one cell and then the other, even if the user location is not changed), we study the number of distinct cells that a user visit instead of the number of cells a user actually cross during a certain period of time, which represent the moving range of the user when he/she is connecting into Mobile Internet and how frequently the user moves across.
There are Formula cells in the area. In this paper, we define four groups of users according to their mobility to represent the user moving activities in the week. Define Formula the number of a cell user access in a week. Accordingly, we define user groups in terms of Formula:
Non mobility users, if
FormulaTeX Source
$$egin{equation*} U_{c}=1 end{equation*}$$
Low mobility users, if
FormulaTeX Source
Normal mobility users, if
FormulaTeX Source
High mobility users, if
FormulaTeX Source
Fig. 6(a) shows CDF of the number of distinct cells users visited during a day. It suggests that CDF of the number of distinct cells for each day is nearly the same. On December 28, 2013, around 35% of the users visited only one cell (non mobility users) and 90% of users visit less than 10 cells in a day. In addition, only 1% of users visited more than 24 cells and users with the highest mobility visited 249 cells. Fig. 6(b) is the CDF of number of distinct cells users visited in a week. The proportion of non mobility users is about 12%, and 50% of users access more than 10 cells. The most active users visited 917 cells in the week. The proportion of high mobility users is about 10% in the whole week. Basically, the more people across different cells, the more hand off activities he will generate, and therefore the more radio resource he will consume.