This study comprehensively reviews research trends from 2015 to 2025 concerning the prediction of challenging behaviors in individuals with severe developmental disabilities through the integration of wearable biosignals and machine learning techniques. The primary objective is to establish a robust theoretical foundation for the practical application of these technologies within the Korean context. Adhering strictly to the PRISMA 2020 guidelines, a systematic literature search was conducted, resulting in the selection of eight international studies published over the last decade. The subsequent rigorous analysis concentrated on critical elements, including the diverse types of biosignals utilized, the specific machine learning algorithms employed, and their reported prediction performance metrics. Key findings from this analysis confirm that various biosignals, notably heart rate, motion data, and sleep patterns, serve as significant and reliable variables for effectively predicting challenging behaviors. A particularly noteworthy insight is that person-dependent models, which are specifically trained to learn and adapt to an individual's unique physiological characteristics, consistently demonstrated higher prediction accuracy, achieving a mean Area Under the Curve (AUC) of 0.84, compared to more generalized global models. Moreover, the systematic review highlighted long-term sleep data as an exceptionally crucial factor contributing to the robustness of challenging behavior prediction. In conclusion, this research firmly establishes that the convergence of wearable biosignals and machine learning possesses substantial potential to fundamentally transform the current intervention paradigm from a reactive to a proactive approach. For the successful and effective domestic implementation of this advanced technology, the study emphasizes the critical necessity for developing highly personalized predictive models, alongside careful consideration of ethical implications and fostering strong multidisciplinary collaboration. Future research efforts should be specifically tailored to address the unique cultural and practical contexts within Korea.

Wearable, Biosignal, Physiological Signal, Machine Learning, Challenging Behavior