An approach to detect smartphone addiction through activity recognition and app usage behaviour

Abstract

The widespread use of smartphones has raised concerns about problematic smartphone use or addiction, which has become a significant issue in today’s society. Despite the recognition of this research area, detecting smartphone addiction remains a challenge. Therefore, it is crucial to identify the primary causes of smartphone addiction and understand how individuals’ lifestyles contribute to this behavior. Most of the methods in research area are self assessment based and detected via different addiction scales. Moreover, in previous studies daily human activities was never considered as a factor in problematic smartphone use. This study aims to explore a new approach in detecting excessive smartphone usage by considering the impact of sensor based daily activities and smartphone app usage. By examining addictive characteristics of smartphone usage and clustering them based on various independent variables, we sought to determine smartphone addiction and investigate the influence of daily activities. To collect reliable and accurate data, we utilized apps for seven days to capture information on the participants’ smartphone usage. Leveraging sensor data and LSTM models, we identified participants’ activities and correlated them with daily app usage duration to detect smartphone addiction using clustering methods such as K-Means and K-Medoids. Our analysis revealed that around 28% participants showed addicted behaviour. To validate these findings, we compared our result with survey results using diverse evaluation metrics (RI,FMI), which exhibited 87% accuracy.