Machine learning facilitates the development of interconnecting layers for perovskite/silicon heterojunction tandem solar cells with proof-of-concept efficiency > 38%

• Machine learning models (Lasso, random forest, multilayer perceptron) were applied to analyze interconnecting layer parameters in perovskite/silicon tandem solar cells. • Key experimental parameters influencing interconnecting layer performance were identified through feature importance analysis. • The optimized interconnecting layer achieved a proof-of-concept efficiency of 38.17%. • The study provides a data-driven approach to accelerate the development of high-efficiency tandem solar cells.