Mining User Attributes Using APP Related Data Of Smartphones

The Mobile Internet has been involved into rapid development and sped up its steps into the "Age of BigData" due to the inspiration that people prefer to access information and service of the internet from everytime and everywhere.As the subject of the Mobile Internet,smartphones achieve continuous innovation to adjust to the explosive demands of users for a long time.Users frequently interact with smartphones which generate large scales of personalized logs.These data are associated with user behaviors tightly,so they implicit rich information about user attibutes,such as personalized information,user habbits and user interests.We can understand users comprehensively and objectively and finally help to improve device,service and application by mining user attribtues.Recently,there have been more people doing research on using mobile data to mine user attributes.However,researchers concentrate on applying location information like cell tower,GPS,Wifi spot.etc.or social information like contact,Call Detail Record.etc.to mine user attributes mostly.The APPs on smartphones provide entrances to kinds of services to users.APP related data can also reflect user basis attributes,user habbits,user preference and user life styles.etc.which provides us a new chance to understand users better.APP installation list(APPList)and APP usage sequence(APPUsage)contain abundant information and can be captured from smartphones easily.In this paper,we concentrate on using APP related data to mine user attributes.The contribution includes:1.Propose a solution for user attributes mining problem based on APP related data and implement frameworks to APPList and APPUsage.2.Explain details of data preprocessing which makes better understanding of these two datasets.Design user representation methods from a number of perspectives and ex-plain the meaning of different features.Try several classification methods based on the designed user representation methods.3.Compare the ability of user attributs mining from different app related data.