Modeling the Detection of Internet Addiction Disorder Using Bayesian Networks

Today’s life revolves around online activities and has made internet a necessity. The unregulated excessive use of internet interferes in daily life and may lead to addiction to it.  In literature, Internet Addiction Disorder (IAD) is defined as compulsive online behaviour which dominates and becomes organizing principle of one’s life impairing physical, emotional and social well-being. The common conventional assessment tools available to detect IAD are Internet Addiction Test (IAT), Compulsive Internet Use Scale (CIUS), Problematic Internet Use Questionnaire (PIUQ).  Recently researchers are adopting Machine Learning (ML) to automate the detection of IAD. This study proposes to develop graphical interface IAD models by harnessing the significant features of Bayesian Networks (BN), a powerful ML technique. The models are constructed with real data sets collected online through IAT, CIUS and PIUQ questionnaires as domain knowledge and we propose these models as IAT-BN, CIUS-BN and PIUQ-BN respectively. The graphical presentations of models provide an effective visualization of causes and symptoms of IAD and facilitate in the better explanation of its occurrence. The promising experimental results show   superior performance of the IAT-BN model with 100% accuracy followed by CIUS-BN as 95% and PIUQ-BN model with 90% accuracy. The study also highlights the important factors which when controlled may substantially reduce the risk of IAD. The models can be an efficient psychiatric decision support system in predicting the occurrence of IAD, risk assessment and management.