Predicting social media addiction fromInstagram profiles: A data mining approach

In this short paper, we describe an application of data mining techniques to predict Instagram users’ addiction from a set of features related to (i) Instagram captions extracted from photos, videos, comments, and stories, and Instagram indicators such as number of followers and following, blocked and closed friends, and frequency of use. We first applied text mining to explore and describe the main contents of Instagram captions. Next, we used a set of non parametric models and ensemble methods to predict Instagram addiction as measured by the Instagram addiction scale [1]. Models were compared via cross-validation using test and training (random) sets from the original dataset. Results showed that Instagram addiction is mainly predicted by the overall time spent on Instagram, writing stories and comments, and number of followers. Moreover, the results suggest that Instagram users made use of photos/videos and stories/comments differently, with the latter being mostly related to emoticons, experiences, and relationships with other users.