Noise Reduction of Mobile Sensors Data in the Prediction of Demographic Attributes

Abstract

In this paper we attempt demonstrate how we can use smartphone sensor data effectively for predicting gender. We specifically focus on sensor data that is assumed to inflict minimal risk to other applications, the system, or the user: Installed Applications, Network Traffic Amount, and Accelerometer readings. We propose several simple heuristics for pre-processing the data and for noise reduction which eventually results in improved accuracy in predicting gender.