Bibliometric Analysis of Deep Learning Research Trends in Primary Education in Indonesia in 2025

Abstract

Learning in elementary schools in Indonesia still tends to emphasize memorization, so that students' deep understanding and critical thinking skills have not developed optimally. This condition calls for alternative learning approaches, one of which is deep learning, which encourages the interconnection of concepts and the application of knowledge in real contexts. This study aims to analyze trends in deep learning research in elementary education in Indonesia throughout 2025 using bibliometric methods. Data were collected from Google Scholar through Publish or Perish using the keywords “deep learning” AND “elementary education” OR “primary school,” limited to publications in 2025. A total of 38 articles that met the criteria were analyzed using Mendeley for metadata processing and VOSviewer for keyword mapping. The results showed that the topic of deep learning was mostly associated with literacy, numeracy, the Merdeka Curriculum, and the integration of digital media. Network, overlay, and density visualizations show that research is increasingly oriented towards strengthening basic competencies through contextual learning strategies. These findings indicate a shift in the direction of research from simply utilizing technology to a more meaningful pedagogical approach for elementary school students. This study recommends the need for further research on the practical implementation of deep learning in the classroom, the development of applied learning strategies, and the expansion of cross-time and cross-country analysis to make the results more comprehensive and support the improvement of basic education quality in Indonesia.