Classification of Bullying-Related Web Documents: An Ecological Systems and Machine Learning Approach

The current study examines bullying-related words in web documents by using social big data and machine learning techniques based on the ecological systems theory. A big data approach is ideal for the study of bullying, as the two terms correspond in their primary features. This study asks the following research questions: First, can social big data be classified into bullying perpetration and victimization documents? Second, which ecological system is the most important in predicting documents related to bullying perpetration and victimization? For this study, we collected documents related to bullying-related documents by using online channels in South Korea and classified main keywords and topics through text mining and opinion mining based on the ecological systems theoretical framework. Machine learning analyses were conducted to present a prediction model of perpetration and victimization documents based on ecological factors. Accordingly, the results indicate all areas of the ecological system are important for understanding the bullying phenomena; moreover, they provide insight into the key areas for bullying intervention.