Identifying age group and gender based on activities on touchscreen

Predicting users' age vis-a-vis gender is imperative in the present days. This study identifies a method and analyses the performance of that method for predicting the users' age (<= 18/18+) and their gender(male/female) using the pattern developed while typing on a touchscreen of a smartphone attached with rotational sensors. These sensors of a smartphone can produce rich data (including timing, rotation, forces) to present how the users of a specific class interact with the phone. A machine learning (ML) technique, Extreme Gradient Boosting (XGBoost), has been applied to develop the models. Empirical results using the dataset gathered from 92 users indicate the innovation of the work resides in the matter that simply the touchscreen pattern can determine user's traits with reasonable accuracy. Further, the study points a futuristic view that if we fuse the sensors pattern within the timing factors, then much more improved results could be gained.