Deep Learning Approach in Predicting Personal Traits Based on the Way User Type on Touchscreen

Abstract

The aim of this paper is to explore the possibility of revealing two common personal traits: age group (<18/18+) and gender (male/female) by measuring and analyzing the way of typing a simple daily-used text on the touchscreen. Deep learning method has been used to develop the model and LOUOCV (Leave-one-user-out cross-validation) has been used to evaluate the effectiveness of our model on the dataset created by 92 volunteers through a web application. Our method outperforms the entire model developed so far as per our knowledge. Accuracy more than 98% in identifying age group and more than 88% in identifying gender have been observed. Automatic age group and gender recognition could be used in a large number of possible application areas such as human-computer interaction, digital forensics, age-specific access mechanism, and targeted advertisement. The way user type on touchscreen concerns with the different issues in user identity verification has been studied well in literature but deep learning approach in revealing the age group and gender based on the typing pattern particularly on a touchscreen is an original idea. Analysis of data, results, and inferences here give a primary account toward achievability to the goal with ample scope for further work in this domain.