Machine Learning-based Prediction of Power Conversion Efficiency for Organic Photovoltaic Materials

Firstly, a mixed neural network model integrating graph convolutional network (GCN) and graph attention network (GAT) is built based on the OPV material data collected from the public literatures. It accurately predictes the power conversion efficiency (PCE) of organic photovoltaic materials. Secondly, the model is fine–tuned on the Harvard organic photovoltaics dataset HOPV15 through the transfer learning technique. Compared with the existing methods, this model improves the prediction accuracy. This method can be widely applied to the rapid preliminary screening of new OPV materials and accelerate the research and development process of new OPV materials.