Which Keywords Grouping and Novelty Trends are Driving Deep Learning Research in Mathematics Education?

Abstract

This study aims to analyze the development of deep learning research in mathematics education using a bibliometric approach. Bibliometric methods are used to evaluate and map scientific literature through statistical analysis of publications, citations, keywords, and collaboration networks. Data were collected from the Scopus database using specific keyword combinations, then filtered using the PRISMA method, resulting in 72 relevant documents for analysis. Data analysis was performed using the R programming language to identify publication trends over time, and VOSviewer software to perform keyword clustering and keyword novelty analysis to uncover thematic clusters and emerging research topics.  The analysis concludes that research on deep learning in mathematics education has experienced significant growth, particularly in recent years, with a sharp increase in the number of publications in 2024 indicating growing interest and research focus in this field. Through keyword clustering, four main themes were identified: computational models for problem-solving, predictive modeling and data analysis, intelligent systems for academic achievement, and curriculum strategies and teaching methods, reflecting the diversity of approaches and applications of deep learning in mathematics education. Furthermore, keyword novelty analysis indicates promising new research opportunities, particularly in the concepts of “Contrastive Learning” and “Adversarial Machine Learning”, which are not yet widely applied but have great potential to improve learning personalization and the robustness of AI-based learning systems. Thus, this trend underscores the importance of bibliometric analysis to map developments, identify research opportunities, and guide the future application of deep learning in mathematics education.