Research On The Distributions Of Main Data Services Of Mobile Internet

Mobile network has become new impetus of the internet due to3G popularity, wireless network development and innovative mobile applications. Web browsing, communication, videos, shopping on line, payment and reading by mobile devices have become important parts of human life. Therefore, analyzing the records of mobile Internet and digging the patterns of human online behavior are vital important for understanding spatial and temporal characters of human behavior.One aspect of human dynamics is studying the temporal characteristics of human behavior, which has attracted scholars’great interests. Many scholars dig temporal patterns of human behavior by analyzing various datasets of human activities, including web browsing, communications by Internet, task executions and other records. They establish models of human dynamics to reveal the origin of heavy-tailed characteristics. Many papers believe that inter-event time distribution decays power-like, which did not accepted by all the scholars. In the paper, we research the temporal models of mobile users based on real Internet records.We cooperate with one of Chinese major operators and analyze the dataset coving1.5million mobile internet users in two cities. In the paper, we analyze the patterns of main data services. Firstly, we observe that subscribers’ online behaviors show a high degree of burstiness, memory, heterogeneity and daily seasonality. Secondly, we study the heavy-tailed distributions of times and traffic generated by individuals. Also we analyze the patterns of up and down traffic in various services. Finally, we classify users by their preferences for Internet services based on Bayesian classification methods. We analyze temporal patterns of users’ accessing web browsing, instant messaging and streaming and model interval time probability density distributions at aggregated level and individual level.