Data-driven visualization of the dynamics of machine learning in materials research

Abstract

The intricate interplay between material structure and properties lies at the heart of modern materials research. Understanding and manipulating this relationship is essential for the development of advanced materials with tailored properties for a wide range of applications. Machine learning (ML) has been intensively employed for prediction purposes. This trend of research into new insights, techniques, and research paradigms is gaining great popularity and demonstrating its promising potential for materials research. This study aims to conduct a bibliometric analysis of ML applications in materials, offering researchers, particularly those in the green energy sector, insights to incorporate into their future research plans. Here, the dataset was retrieved from the Web of Science Core Collection and the earliest related publication was recorded in 1998. Metrics based on retrieved data were extracted, including publication evaluations, countries, journals, and authors. Keywords temporal variations and citation-based scientific landscapes were constructed. The findings underscore the embryonic nature of machine learning's deployment in materials research but also highlight its significance as an emerging field that has captured the attention of scholars across multiple domains. Specifically, ongoing research efforts are directed towards optimizing ML models and algorithms, as well as refining data handling techniques to glean insights into complex structure-property relationships. The findings will provide novices with a data-driven visualization summary about the dynamics of this field, and its inspiration to environmental sustainability, and benefit a wide range of stakeholders to enhance their informed decisions on research funding and policy.