BEACON-AI: A CNN- BiLSTM -Attention Framework for Real-Time Multimodal Student Behavior Analysis and Academic Early Warning
Abstract
Monitoring students' behavior and engagement in real time is essential to improve learning outcomes and reduce academic risk. As a result of their reliance on static academic records and manual observations, traditional early warning systems often fail to identify behavioral markers predictive of declining involvement. BEACON-AI is a real-time deep learning platform proposed in this work for multimodal student behavior analysis and academic risk prediction. The system utilizes physiological, behavioral, and environmental data. These data are obtained from 30 days of wearable Internet of Things sensor records on Kaggle. These recordings include facial expressions, posture, heart rate, movement, and classroom context. To capture temporal and contextual patterns, the data preprocessing included normalization, one-hot encoding, and segmentation into sliding time-series windows, among other steps. BEACON-AI uses a CNN-BiLSTM-Attention hybrid model to forecast attention, engagement, and inactivity at multiple levels. With accuracy rates of 94.2% for attention detection, 92.5% for engagement categorization, and over 85% for early disengagement prediction, the system outperformed baseline models primarily used in academic settings. To demonstrate a high level of dependability in identifying behavioral changes, task-wise accuracy, recall, and F1 Scores were specifically measured. It is possible to comprehend the results of the attention mechanism since it highlights the most critical behavioral contributions. The overall value of combining multimodal behavioral and physiological data for educational interventions is demonstrated by BEACON-AI, which provides real-time, scalable behavioral monitoring and proactive academic early warning.References
Bagunaid, W., Chilamkurti, N., Shahraki, A. S., & Bamashmos, S. (2024). Visual data and pattern analysis for smart education: a robust DRL-based early warning system for student performance prediction. Future Internet, 16(6), 206.
Deng, J., Huang, X., & Ren, X. (2024). A multidimensional analysis of self-esteem and individualism: A deep learning-based model for predicting elementary school students’ academic performance. Measurement: Sensors, 33, 101147.
Wang, Z., Li, L., Zeng, C., Dong, S., & Sun, J. (2025). SLBDetection-Net: Towards closed-set and open-set student learning behavior detection in smart classroom of K-12 education. Expert Systems with Applications, 260, 125392.
Gao, Y. (2025). Deep learning-based strategies for evaluating and enhancing university teaching quality. Computers and Education: Artificial Intelligence, 8, 100362.
Jin, J. (2025). Research and Construction of Student Management Platform for Special Needs Students with Decision Tree Model and Big Data Technology. Systems and Soft Computing, 200310.
Jayaprakash, D., & Kanimozhiselvi, C. S. (2024). Multinomial logistic regression method for early detection of autism spectrum disorders. Measurement: Sensors, 33, 101125.
Zaibi, T., & Bezine, H. (2024). Early detection of learning disabilities through handwriting analysis and machine learning. Procedia Computer Science, 246, 3702-3712.
Mumenin, N., Hossain, A. K., Hossain, M. A., Debnath, P. P., Della, M. N., Rashed, M. M. H., ... & Hossain, M. S. (2024). Screening depression among university students utilizing GHQ-12 and machine learning. Heliyon, 10(17).
Sulaiman, M. H., & Mustaffa, Z. (2024). Enhancing wind power forecasting accuracy with hybrid deep learning and teaching-learning-based optimization. Cleaner Energy Systems, 9, 100139.
Kim, D., Lee, K., Jeong, S., Song, M., Kim, B., Park, J., & Heo, T. Y. (2024). Real-time chlorophyll-a forecasting using machine learning framework with dimension reduction and hyperspectral data. Environmental Research, 262, 119823.
Uddin, M. A., Talukder, M. A., Uzzaman, M. S., Debnath, C., Chanda, M., Paul, S., ... & Aryal, S. (2024). Deep learning-based human activity recognition using CNN, ConvLSTM, and LRCN. International Journal of Cognitive Computing in Engineering, 5, 259-268.
Chen, Y., Sun, J., Chen, Y., Li, E., Lu, J., Tang, H., ... & Sun, B. (2025). Machine learning-based model for acute asthma exacerbation detection using routine blood parameters. World Allergy Organization Journal, 18(7), 101074.
Li, L., Guo, D., Shi, C., & Zheng, Y. (2025). The predictive role of sedentary behavior and physical activity on adolescent depressive symptoms: A machine learning approach. Journal of Affective Disorders, 378, 81-89.
Yousaf, M., Farhan, M., Saeed, Y., Iqbal, M. J., Ullah, F., & Srivastava, G. (2024). Enhancing driver attention and road safety through EEG-informed deep reinforcement learning and soft computing. Applied Soft Computing, 167, 112320.
Rajaram, Sampath. “A Model for Real-Time Heart Condition Prediction Based on Frequency Pattern Mining and Deep Neural Networks.”, PatternIQ Mining.2024, (01)1, 1-11.
Zhou, Y., Wang, J., & Zhang, J. (2024). A Multimodal Image Recognition System for Student Behavior Analysis in Smart Classrooms in Universities. Traitement du Signal, 41(6).
Yusuf, A., Noor, N. M., & Bello, S. (2024). Using multimodal learning analytics to model students’ learning behavior in animated programming classroom. Education and Information Technologies, 29(6), 6947-6990.
Zhou, Q., Suraworachet, W., & Cukurova, M. (2024). Detecting non-verbal speech and gaze behaviours with multimodal data and computer vision to interpret effective collaborative learning interactions. Education and information technologies, 29(1), 1071-1098.
Li, H., Mao, Y., Xu, Y., Tu, K., Zhang, H., Gu, R., & Sun, Q. (2025). Rapid detection of the viability of naturally aged maize seeds using multimodal data fusion and explainable deep learning techniques. Food Chemistry, 478, 143692.
Ji, X., Sun, L., & Huang, K. (2025). The construction and implementation direction of personalized learning model based on multimodal data fusion in the context of intelligent education. Cognitive Systems Research, 101379.
Xin, R., Pan, J., Yang, F., Yan, X., Ai, B., & Zhang, Q. (2024). Graph deep learning recognition of port ship behavior patterns from a network approach. Ocean Engineering, 305, 117921.
Darban, Z. Z., Webb, G. I., Pan, S., Aggarwal, C. C., & Salehi, M. (2025). CARLA: Self-supervised contrastive representation learning for time series anomaly detection. Pattern Recognition, 157, 110874.
Fan, S., He, T., Li, W., Zeng, C., Chen, P., Chen, L., & Shu, J. (2024). Machine learning-based classification of quality grades for concrete vibration behaviour. Automation in Construction, 167, 105694.
Ye, X., Kevin, I., & Wang, K. (2024). Deep generative domain adaptation with temporal relation attention mechanism for cross-user activity recognition. Pattern Recognition, 156, 110811.
Xu, Z., Cai, B., Yan, L., Pang, X., & Gao, K. (2024). Statistical analysis of metastable pitting behavior of 2024 aluminum alloy based on deep learning. Corrosion Science, 233, 112077.
Lam, P. X., Mai, P. Q. H., Nguyen, Q. H., Pham, T., Nguyen, T. H. H., & Nguyen, T. H. (2024). Enhancing educational evaluation through predictive student assessment modeling. Computers and Education: Artificial Intelligence, 6, 100244.
Embarak, O. H., & Hawarna, S. (2024). Automated AI-driven system for early detection of at-risk students. Procedia Computer Science, 231, 151-160.
Zhao, M., Song, Y., Huang, H., & Kim, E. H. (2025). Attention-based fuzzy neural networks designed for early warning of financial crises of listed companies. Information Sciences, 686, 121374.
Jeong, H., Park, S., Choi, B., Yu, C. S., Hong, J. Y., Jeong, T. Y., & Cho, K. H. (2024). Machine learning-based water quality prediction using octennial in-situ Daphnia magna biological early warning system data. Journal of Hazardous Materials, 465, 133196.
Weng, P., Tian, Y., Zhou, H., Zheng, Y., & Jiang, Y. (2024). Saltwater intrusion early warning in Pearl river Delta based on the temporal clustering method. Journal of Environmental Management, 349, 119443.
https://www.kaggle.com/datasets/ziya07/student-behavior-monitoring-datas
DOI:
https://doi.org/10.31449/inf.v50i10.10917Downloads
Published
How to Cite
Issue
Section
License
Authors retain copyright in their work. By submitting to and publishing with Informatica, authors grant the publisher (Slovene Society Informatika) the non-exclusive right to publish, reproduce, and distribute the article and to identify itself as the original publisher.
All articles are published under the Creative Commons Attribution license CC BY 3.0. Under this license, others may share and adapt the work for any purpose, provided appropriate credit is given and changes (if any) are indicated.
Authors may deposit and share the submitted version, accepted manuscript, and published version, provided the original publication in Informatica is properly cited.
