These findings were supported by the results of our analytic
modeling and the analysis of our interview data. Regression
analysis showed that the key components of smartphone
overuse, such as interference, withdrawal, and tolerance,
were closely related to usage features. In addition, we used
machine learning techniques to test the predictive power of
the usage features and found that smartphone overuse could
be classified accurately based on these features with an Fscore
of 0.87. The interview results provided more detailed
examples of problematic usage behavior, such as limited selfcontrol
when consuming online content (e.g., aimlessly following
web/Facebook links while in bed). Our key findings
support the existing theories related to technological addictions
[11, 17, 10]. In particular, repeated consumption of online
content may lead to addictive behaviors, and problematic
behaviors depend mainly on specific functions rather than the
volume of usage