Harnessing the leading edge: machine learning ventures in chemistry and materials science

Abstract

The widespread application of machine learning (ML) is profoundly transforming traditional research methods in materials science and chemistry, bringing new opportunities while also posing significant challenges and risks. Improper use of ML methods can lead to biased and misleading research outcomes. This review outlines the application processes of ML in the fields of materials science and chemistry, providing an in-depth analysis of potential issues at each stage with case studies, including data management, model construction, evaluation, and shared risks in data reporting. We emphasize the necessity of standardized use of ML and highlight the current crises faced in ML applications in scientific research. This review also summarizes a series of strategies to ensure the reliability and scientific validity of research results. It aims to offer practical guidance to researchers, helping them leverage the advantages of ML while applying these tools in a scientifically sound and compliant manner, avoiding common pitfalls, and promoting more rigorous research practices in materials science and chemistry.