An AIoT-Based Interactive Literacy Game Using Micro:bit and Ensemble k-NN–Decision Tree for STEAM Learning

Hayatunnufus (University of Sumatera Utara)
Nirmala Aizya Agatha Silalahi (University of Sumatera Utara)

Data Science: Journal of Computing and Applied Informatics
Vol. 10 No. 2 (2026): Data Science: Journal of Computing and Applied Informatics (JoCAI) In Press

Low literacy performance among Indonesian students, as indicated by PISA result, highlights the need for innovative learning approaches. This study proposes an interactive literacy game integrating Micro:bit and machine learning within an AIoT framework for STEAM education. The system uses accelerometer data to recognize hand gestures (left, right, up) as multiple-choice answers. A dataset of 61,148 gesture samples was collected and processed. An ensemble model combining k-Nearest Neighbor (k-NN) and Decision Tree algorithms achieved a classification accuracy of 90.25%. Real-time implementation with elementary students (n=19) yielded a gesture recognition accuracy of 95.03%. User testing showed high engagement (8.5/10) and positive learning impact (8.9/10), demonstrating the system’s effectiveness as a lightweight, interactive tool for literacy education.