In order to reveal the internal logic of the co-evolution of digital economy industrial ecology and artificial intelligence, an analysis framework of the co-development mechanism driven by graph neural network was constructed. Starting from the enterprise, platform, technology supply and regional unit, this paper designed a heterogeneous collaborative network structure, established the linkage mechanism of hierarchical representation learning, ecological perception and intelligent decision-making, and carried out effect test combined with multi-regional samples. The results show that the model performs better in collaborative identification and optimal configuration, with the Accuracy reaching 91.7%, Macro-F1 reaching 90.3%, and AUC reaching 93.6%. The collaborative efficiency of typical areas, the degree of cross-subject interaction and the matching rate of artificial intelligence are improved. The research shows that graph neural network can effectively depict multi-agent association, cross-layer propagation and regional differences in the industrial ecology of digital economy.
