Interpretable Artificial Neural Networks for Band Gap Prediction in 2D Hybrid Organic–Inorganic Perovskites

Two‐dimensional hybrid organic–inorganic perovskites (2D‐HOIPs) possess remarkable photoelectric properties, including strong light absorption, high electrical conductivity, and long carrier lifetimes, making them promising candidates for optoelectronic applications. This study aims to accurately predict their band gaps using machine learning (ML) to identify high‐performance 2D‐HOIPs. A total of 354 data points are collected from the HHPMDB database, and 32 compositional and structural features are selected via recursive feature elimination with fivefold cross‐validation. An Artificial Neural Network (ANN) model is developed, achieving an excellent predictive performance with an R 2 of 0.926. Shapley Additive Explanations (SHAP) analysis is employed to interpret feature contributions to the band gap. We compared the predicted values from our models with those calculated using Generalized Gradient Approximation (GGA), ensuring an error range of approximately 0.2 eV, thereby confirming the accuracy of our models. Additionally, comparisons between Perdew–Burke–Ernzerhof (PBE) and High Local Exchange 2016 (HLE16) band gaps further confirmed model accuracy. This approach enables rapid and cost‐effective prediction of the 2D‐HOIP band gap.