A streaming PAC-learning framework for real-time detection of gaming addiction and toxic behaviour among Indian adolescents

This dataset contains behavioural, demographic, and communication-based records of 5,000 Indian adolescents (ages 12–18) collected across the years 2020–2024 for the study “A Streaming PAC-Learning Framework for Real-Time Detection of Gaming Addiction and Toxic Behaviour.” It includes multimodal features such as average session duration, in-game action intensity, and chat toxicity scores, along with volatility indicators that capture behavioural drift over time.

The dataset is synthetic but statistically modelled to replicate real-world adolescent gaming patterns, ensuring realistic correlations between gaming intensity and toxicity. Year-wise behaviour patterns match published trends, reflecting increasing session duration, aggression indicators, and rising toxicity levels among adolescents. Each record also includes a model-generated toxicity label (0/1) used for supervised learning and evaluation of the proposed streaming PAC-learning framework.

This dataset supports research on online gaming behaviour, addiction detection, early-warning systems, streaming analytics, behavioural modelling, sentiment analysis, and machine-learning-based risk prediction. It is anonymized, contains no personally identifiable information, and is suitable for academic replication, algorithm testing, and simulation of real-time behavioural monitoring systems.