<![CDATA[This paper aims to explore the concepts and themes emerging in the literature related to "Deep Learning in Mathematics Learning" in order to understand the direction of development and current trends in the field. To achieve this goal, the study uses the Automatic Systematic Literature Review (SLR) method with the help of Scopus AI, which allows for the automatic identification of concepts and themes through the visualization of concept maps and emerging themes. The database selection focused solely on Scopus due to its high reputation and extensive coverage of high-quality international journals. The keyword used is "deep learning in mathematics learning" with a publication time limit from 2003 to 2025, thus covering early developments to the latest trends. This approach allows for systematic and efficient literature mapping without having to manually review all documents. The analysis reveals that the topic of "Deep Learning in Mathematics Learning" encompasses several emerging themes, including student performance prediction, AI integration in mathematics education, and the adoption of innovative pedagogical practices. Based on the concept map visualization, three main research directions are identified: Learning Environment, Techniques, and Applications. The theme of student performance prediction highlights the use of neural network models such as CNNs and LSTMs to analyze key factors influencing academic outcomes. Meanwhile, AI integration focuses on the development of adaptive learning platforms that personalize instruction and enhance learning effectiveness. Innovative pedagogical practices, including the use of extended reality and machine learning, aim to create immersive and interactive learning experiences. Overall, these findings underscore the significant potential of deep learning to transform mathematics education through intelligent, adaptive, and student-centered approaches.]]>
