移动互联网中基于用户行为的机会连接预测

Opportunistic link prediction based on user behavior in mobile Internet

为提高用户机会连接预测的准确率,进而提升网络服务性能和效率,通过对我国浙江省某市运营商的真实数据集进行模拟实验,研究了用户行为特征与机会连接的相关性.通过分析相遇用户对和非相遇用户对的热点位置分布,空间相似度分布得出用户移动性特征对用户机会连接行为的影响,并在搭建复杂网络模型的基础上采用随机森林算法,综合考虑了网络结构特征、用户移动性特征、用户上网行为特征,对移动互联网中个体在未来一天中某个时间段是否会发生机会连接进行了预测,结果表明,与传统预测方法相比,加入用户移动性特征、用户上网行为特征后的复杂网络模型具有更高的准确率与召回率,利用此模型可更好地提升网络服务性能和效率.

In order to improve the accuracy of user opportunity link prediction,and then improve the performance and efficiency of network services,this paper studies the correlation between user behavior characteristics and opportunistic link through simulation experiments on a real data set of operators in Zhejiang Province.In this paper,by analyzing the distribution of hot spots and spatial similarity between encounter users and non-encounter users,we get the influence of user mobility characteristics on user’s opportunistic link behavior.On the basis of building complex network model,we adopt random forest algorithm,which takes into account the network structure characteristics,user mobility characteristics and the characteristics of user’s Internet access behavior.The results show that the complex network model with user’s Internet access behavior characteristics and user mobility characteristics has higher accuracy and recall rate than the traditional prediction method.This model can better enhance the performance and efficiency of network services.