Demographic prediction is an important component of user profile modeling. Accurate prediction of users’ demographics can help many applications, ranging from web search to behavior targeting.
In this project, we focus on how to predict mobile Internet users’ demographics of gender, age and college of students on campus, based on spatial-cyber-semantic analysis (including user trajectory and pattern extraction, user online web page analysis and semantic mining on user search keywords), using dataset collected from SJTU campus Wi-Fi networks.
The prediction accuracy of the ensemble model is around 85%.