基于移动感知数据的用户画像系统

User Portrait System Based on Mobile Sensing Data

用户画像是一个人的虚拟表示,它是基于一系列数据的模型.基于手机感知数据从年龄、性别和人格特征三方面构建用户画像.通过使用手机中的传感器和事件监听器来采集滑屏解锁场景、手机基本信息、应用程序使用情况和屏幕状态场景中的数据.此外,随机森林回归和随机森林分类模型分别用于估计年龄和检测用户的性别.支持向量回归机(support vector regression,SVR)算法用于识别人格特征.通过84个用户进行实验来评估该模型.实验结果表明,我们的方法在年龄估计中的均方根误差为4.369,在性别检测中实现了91.70%的精度.对于人格特征的识别,开放性、尽责性、外倾性,宜人性和神经质的均方根误差分别为0.290、0.351、0.465、0.302和0.452.

A person was represented digitally as a user portrait,which was a model based on a series of data.User portrait has been constructed from age,gender and personality traits based on mobile phone sensory data.Specifically,data has been captured in unlocking screen,phone basic information,app usage and screen status scenes by using ubiquitous sensors and event listeners available in mobile phone.Furthermore,random forest regression and random forest classification models were separately used to estimate the age and detect the gender of the user.SVR algorithm was applied to identify personality traits.Model has been evaluated through real-life experiments conducted by 84 phone users totally.Experimental results showed that our approaches achieved the RSME of 4.369 in age estimation and 91.70%precision in gender detection.For personality traits identification,the RMSEs of openness,conscientiousness,extroversion,agreeableness and neuroticism were 0.290,0.351,0.465,0.302 and 0.452,respectively.