Community Division of Multi-factor Coupled Complex Network Based on Graph Neural Network

Complex electromechanical systems in the process industry consist of numerous mutually coupled monitoring variables, forming a dynamic coupling relation network that reflects the operational state of the system. Aiming at the above problems, a method of community division of multi-factor coupled complex network based on graph neural network is proposed. This method firstly establishes a dynamic directed weighted dynamic network based on the coupled Granger causality measure analysis, and then conducts graph neural network training, and finally obtains the result of coupling network community division. The results show that the method proposed in this paper has more obvious advantages in mining network structure information than traditional methods such as K-means and BGLL, and the effectiveness of the proposed method is also verified by actual enterprise cases.