Learning-Advisory System Based on Students’ Similarity

The volume and the complexity of data on students’ learning behaviors make it challenging to extract valuable insights. This study aims to build a learning advisory system that identifies low-performing students and offers guidance to improve their academic performance. Here, we propose a method to map students’ learning characteristics into a two-dimensional topological representation using Self-Organizing Maps (SOM) to discover their learning style similarity using m-Ary Tree. We name the proposed method SOM-m-AT. The study utilized data from MONSAKUN, a digital learning framework in Japan focusing on problem-posing for elementary-level mathematics. By mapping the learning characteristics of students, teachers can identify low-performing students and those with better performance in their vicinity on the map. This approach can help educators identify successful learning patterns and suggest these to low-performing students to improve their academic results. Through empirical experiments, the SOM-m-AT method showed how students with similar learning behaviors tend to cluster together on the map, which can help teachers design targeted interventions and provide personalized guidance to low-performing students. By leveraging these patterns, teachers can provide more targeted feedback and support to students struggling with arithmetic story problems. This paper presents the basic framework for the proposed system and reports the results of empirical experiments.