Surface catalyst adsorption energy prediction method based on improved graph neural network

The invention discloses a surface catalyst adsorption energy prediction method based on an improved graph neural network. The method relates to the fields of computational chemistry, deep learning, surface catalyst adsorption energy prediction and the like. The method aims to solve the problems that most of existing graph neural networks ignore effective input feature coding on atoms by utilizing potential spatial information and mainly pay attention to basic connectivity and element features; and the existing graph neural network architecture mostly takes extraction of local neighborhood features of the atomic graph as guidance and ignores the integration problem of global node features. The technical scheme adopted by the invention is as follows: by combining a uniform atomic position smooth overlapping descriptor and a graph weak attention mechanism introduced into a graph network architecture, local space structure information and global node information of an atomic graph are effectively extracted, and t