Machine learning: from perovskite materials to devices

To address the growing need for efficient and sustainable energy solutions, this review focuses on the integration of machine learning (ML) into perovskite materials research, which has become increasingly important due to the materials' tunable structures and excellent optoelectronic properties. Traditional experimental methods struggle with the complexity of perovskite systems, highlighting the need for data-driven approaches. This work systematically summarizes ML applications in perovskite materials and devices, with an emphasis on interpretability, reproducibility, and practical deployment. Key innovations include the use of Shapley Additive explanations (SHAP) to improve model transparency, the adoption of the MatBench benchmark dataset for standardized model assessment, and the promotion of open, interactive ML workflows through platforms like Jupyter Notebook. Furthermore, this review explores the emerging role of large language models in materials science, offering new avenues for knowledge extraction and intelligent analysis. By bridging materials research with modern ML tools, this work provides a timely reference and practical guide for advancing the design and optimization of perovskite optoelectronic devices.