A comparative analysis of some variables affecting tiktok addiction using machine learning methods

This study aimed to examine the relationship between TikTok addiction levels and loneliness, as well as certain demographic and personal variables, using machine learning methods. The study was conducted using data obtained from 592 university students aged 18 and over across Turkey. Designed as a cross-sectional and correlational survey, the study utilised the Demographic and Personal Information Form, the TikTok Addiction Scale (TAS), and the UCLA Loneliness Scale Short Form (ULS-8). The obtained data were converted into a two-level categorical structure, high/low TikTok addiction level, using Two-Step Cluster Analysis. Following clustering, comparative analyses were performed using machine learning algorithms such as CatBoostClassifier, MLPClassifier, HistGradientBoostClassifier, and LightGBM. According to the research results, the CatBoostClassifier method was the model that best predicted TikTok addiction. Findings indicate that the variable most affecting TikTok addiction is 'daily TikTok usage time'. This variable was followed by 'level of loneliness,' 'purpose of TikTok usage,' and 'activity level.' PDP and SHAP visualisation analyses also revealed that TikTok usage time and loneliness variables are significant predictors of TikTok addiction.