移动互联网的业务访问特性

Revealing Service Visit Characteristics in Mobile Internet

随着移动网络的快速发展和智能手机的不断普及,移动互联网的用户规模与日剧增,各类业务应用也层出不穷.为分析移动互联网中各类业务的访问特性,文中首先引入复杂网络的研究方法,建立了一个加权用户-业务二分网络分析模型;然后对现有移动互联网的主要业务应用进行了分类,并利用从某移动通信运营商的互联网网关上采集的2010年和2011年部分数据集,基于该模型从用户访问兴趣、业务点击量特性、业务流量特性、访问关联性等方面分析了移动互联网的业务访问特性,并比较了用户访问行为的变化.结果发现:门户网站、搜索引擎、社交网站和网络文学是用户访问的主要业务类型;用户访问兴趣范围服从指数分布,用户访问的兴趣强度服从幂律分布;点击量大的网站具有流量相似性而点击量小的网站则没有,但当点击量达到一定规模时则会呈现出流量相似性;在访问关联性方面,门户网站和搜索引擎是最容易被关联访问的业务类型,其次是电子邮箱和社交网站.

With the rapid development of mobile networks and popularization of smart phones, the scale of mobile Internet users is explosively growing, and various kinds of services have emerged. To investigate the characteristics of user access of different kinds of services, we set up a weighted user-service bipartite network model. Before using our model to analyze the user access characteristics, we categorize the main services in mobile Internet into twelve types. Then, based on the real data sets respectively collected in 2010 and 2011 from a WAP gateway of one main Mobile Telecom Carrier in Chongqing province, we target to expose the characteristics of service visit styles from four aspects： access interest, clicks, traffic features, and access rele- vance. We have several interesting findings. （a） the top four service categories being visited are the portal, search, online social networks and e book~ （b） the number of service categories that user visited well fits exponential distribution, while the user visiting intensity fits power law~ and （c） the volume of traffics on websites whose clicks are high would show strong self-similarity, but those having low clicks would not. However, when the volume of clicks on a website reaches a certain high level, it would show the self-similarity. Moreover, （d） with respect to access rele- vance, which means the users＇ access transfers among different service categories, we find that the portal and search are the top two kinds of services which tightly correlated with other service categories, followed by the e-mail and online social networks.