摘要
提出了一种基于支持向量机的预测方法,通过分析智能手机应用的使用情况,预测用户的人口统计信息。手机使用行为数据约为5万智能手机用户在3个月期间使用手机应用产生的网络日志文件,包括179 954 181条日志记录。通过对日志记录的主题进行分析,可将179 954 181条日志记录匹配到266个不同的主题。在此基础上,通过将每个用户的人口统计信息与该用户对266个不同主题的访问权重进行关联,可构建训练数据,并代入支持向量机模型进行计算。实验结果表明该方法对用户的性别和年龄预测能够取得良好的预测结果。
A support-vector-machine-based predicting method is presented to predict users＇ demographic information by analyzing the usage of the applications in the smartphones. The smartphone usage data considered in this paper is a network log file, which records smartphone applications usage of 50 000 smartphone users for three months, including 179 954 181 entries. By analyzing the topic of each entry, the 179 954 181 entries can be matched with 266 distinct topics. Based on this result, by correlating the users＇ demographic information with their query weight of such 266 distinct topics, a training data can be constructed and imported to support vector machine model for computation. The results of experiments show that the method proposed in this paper can well predict uses＇ gender and age.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2015年第6期917-920,933,共5页
Journal of University of Electronic Science and Technology of China
基金
国家自然科学基金(61133016
61300191
61370026)
教育部-中国移动科研基金(MCM20121041)
四川省科技支撑计划(2014GZ0106)
中央高校基金(ZYGX2013J003)
关键词
人口统计信息
预测
智能手机应用
支持向量机
demographic information
prediction
smartphone application
support vector machine
