Mining User Attributes Using Large-Scale APP Lists of Smartphones

Prevalence of smartphones is changing people's lifestyle. Mobile applications (abbr. APPs) on a smartphone serve as entries for users to access a wide range of services. What APPs installed on one's smartphone, i.e., APP list, convey lots of information regarding his/her personal attributes, such as gender, occupation, income, and preferences. This paper addresses the discovery of user attributes from an APP list. We develop an attribute-specific representation to describe user characteristics and then model the relationship between an attribute and an APP list. A large-scale real-world data set with APP lists of more than 100 000 smartphones is used for evaluation. Our approach achieves the average equal error rate of 16.4% for 12 predefined user attributes. To our best knowledge, this is the first work to explore mining of user attributes from installed APP lists.