Vegetation Indices for Multidomain Prediction via Deep Learning: A Scientometric Analysis

Deep learning has considerably enhanced vegetation index prediction accuracy and efficiency, leading to a significant increase in academic publications across multiple domains.However, most existing studies concentrate on specific applications, lacking systematic reviews that explore its interdisciplinary knowledge structure and broader multidomain potential.Therefore, in this study, we investigated vegetation indices and deep learning advancements from both technical and scientometric perspectives.We begin by reviewing the application of deep learning in vegetation index prediction, emphasizing the structural features of several mainstream models and their practical performance within integrated frameworks.Then, on the basis of 173 relevant articles collected from the Web of Science database from 2015 to 2025, we employed CiteSpace to construct detailed networks of author co-citations, journal co-citations, keyword co-occurrences, and international collaboration.Through citation burst detection and betweenness centrality analyses, we uncover evolving research trends, key hotspots, and global cooperation patterns in this dynamic interdisciplinary field.The findings highlight growing research interest, with central themes focusing on method optimization, cross-domain application expansion, and strengthened international collaboration.In this study, we provide a new theoretical perspective for the application of deep learning in vegetation index prediction across multiple domains, while also offering fresh insights for future interdisciplinary collaboration and technological innovation directions.