The increasing integration of digital technologies in education has led to the emergence of cyber-physical learning (CPL) environments, in which the maintenance of learner alertness during human-computer interaction (HCI) has been considered essential for enhancing academic performance. Although multimodal learning analytics (MMLA) has been utilized to identify cognitive and emotional states through physiological data, limited attention has been given to real-time assessment of alertness using portable digital devices. In this study, the feasibility of assessing learner alertness through digital biomarkers in a CPL environment was investigated. A total of forty-one participants were involved in the NPFC-Test, an interactive web-based assessment that included meditation, mathematical tasks, memory recall, writing exercises, and audiovisual stimuli. The experiment was conducted at the IFE Experiential Classroom. Physiological signals were collected using a low-cost EEG headband, a medicalgrade smart bracelet, and an RGB camera for facial coding. Five machine learning models-Random Forest, XGBoost, KNN, Support Vector Machine, and Logistic Regression-were trained to classify alertness levels using multimodal input features. The Random Forest model yielded the highest predictive accuracy, with body temperature and electrodermal activity identified as the most relevant indicators. These results demonstrated that affordable and portable biosensors may be effectively applied for real-time monitoring of alertness, thereby supporting more adaptive learning environments. It is recommended that future research be directed towards the refinement of the predictive model, expansion of the dataset, and validation of the approach in diverse academic settings, with the aim of strengthening the application of MMLA as a tool for educational innovation.
