Comparative Study of Feature Selection and Classification for Problematic Mobile Phone Use (PMPU)

This study explores the issue of problem phone use, and the role of emotion regulation and impulsivity in problematic mobile phone use. We also attempt to study this issue from a machine learning perspective. Feature selection and classification form the basis of machine learning. They complement each other to solve many challenging computer vision, prediction, and classification problem. Rarely has this method been experimented in the context of psychological data. The current study presents a unique comparative approach to examine the relationship between problematic mobile phone use, impulsivity and emotional dysregulation using machine learning’s methodology. The data in this study were collected from 209 University undergraduates, of which 72 are males and 137 are females (M = 21.98 years old, SD = 3.873) using an online survey. The survey includes three questionnaires, namely Barratt Impulsiveness Scale (BIS), Difficulties in Emotion Regulation Scale (DERS), and Problematic Mobile Phone Use Questionnaire (PMPUQ). We present and discuss results obtained using five different feature selection and classification algorithms. The feature selection algorithms come to a consensus that the main contributing factors for problem phone use are due to (i) dependency issue, (ii) financial issue, and (iii) dangerous use. The classification results show consistent accuracy over several performance matrices. The classification hit rate is in the range of 88 to 99%. © 2017, Springer International Publishing AG.