Real Time Smartphone Data for Prediction of Nomophobia Severity using Supervised Machine Learning

Authors

Anshika Arora
Department of Computer Science and Engineering Netaji Subhas University of Technology Delhi, India
Pinaki Chakraborty
Department of Computer Science and Engineering Netaji Subhas University of Technology Delhi, India
M.P.S. Bhatia
Department of Computer Science and Engineering Netaji Subhas University of Technology Delhi, India

Synopsis

Excessive use of smartphones throughout the day having dependency on them for social interaction, entertainment and information retrieval may lead users to develop nomophobia. This makes them feel anxious during non-availability of smartphones. This study describes the usefulness of real time smartphone usage data for prediction of nomophobia severity using machine learning. Data is collected from 141 undergraduate students analyzing their perception about their smartphone using the Nomophobia Questionnaire (NMP-Q) and their real time smartphone usage patterns using a purpose-built android application. Supervised machine learning models including Random Forest, Decision Tree, Support Vector Machines, Naïve Bayes and K-Nearest Neighbor are trained using two features sets where the first feature set comprises only the NMP-Q features and the other comprises real time smartphone usage features along with the NMP-Q features. Performance of these models is evaluated using f-measure and area under ROC and It is observed that all the models perform better when provided with smartphone usage features along with the NMP-Q features. Naïve Bayes outperforms other models in prediction of nomophobia achieving a f-measure value of 0.891 and ROC area value of 0.933.

WREC21
Published
September 22, 2021
Online ISSN
2582-3922