Online game addiction level prediction based on day to day life activity using machine learning

Abstract

The growing popularity of gaming in modern society has raised concerns about the possible dangers of addiction. The major goal is to create a predictive model capable of identifying and categorizing different stages of game addiction based on a wide range of factors, such as gaming behavior, psychological qualities, and socioeconomic information. The dataset, obtained from reliable archives such as Kaggle, includes 51 characteristics covering a wide range of topics such as gender, age, academic qualifications, gaming habits, opinions about oneself, emotional states, and social contacts. The focus attribute, 'Addiction Level', stratifies individuals into categories: 'non-addictive','moderately addictive', and 'very addictive'. The process is thorough, beginning with data cleansing and progressing through feature engineering and exploratory data analysis (EDA) to generate important insights into attribute connections, distributions, and trends. Following a thorough examination, machine learning models are trained and tested using a variety of methods, such as Random Forest Classifier, Gradient Boosting classifiers, Voting Classifier, SVM, and Adaboost classifier. Here on those algorithms SVM outperforms the highest accuracy 90.76%. Furthermore, the study intends to provide an extensive predictive model capable of identifying addiction levels based on numerous behavioral, psychological, and demographic gaming features.