Effectuating Supervised Machine Learning Techniques for Multiclass Classification of Problematic Internet and Mobile Usage

The internet has slowly become an inevitable part of every facet of our lives. With the power of the world wide web available at the touch of our fingertips, anything seems possible. But mental health disorders due to prolonged usage of the ever-evolving internet and mobile are also on the rise. Studies show there is a strong correlation between excessive internet usage and depression, lower self-esteem, Attention-Deficit Disorder (ADHD), impulsivity, hyperactivity and so on. In this paper, a system is proposed that classifies a persons' internet/mobile usage into four classes (multi class) which are- Normal, Borderline, Critical and Severe. In collaboration with our institutions' Counsellor and considering previous studies, a non-invasive questionnaire was developed to collect the data. The collected data was used to train some efficient and state-of-the-art machine learning models such as Logistic Regression, Decision Trees, Support Vector Machines (SVM), Xtreme Gradient Boosting (XGBoost), Random Forests and Light Gradient Boosting (LightGBM). The model with the highest accuracy was taken forward to deliver the best possible classification of a user into one of four categories. With thorough training and testing linear SVM with radial basis kernel returned the best accuracy and thus it was chosen to move forward with.