The Complex Network Model for Industrial Data Based on Spearman Correlation Coefficient

Abstract

In this paper, we propose a complex network design method based on Spearman correlation coefficient. According to the various characteristics of complex networks, we believe that we can build a complex network model for industrial systems. And then we take the boiler in the thermal power generation scenario as an example. By using the Spearman correlation coefficient and the SVR (Support Vector Regression) model in machine learning, we successfully built a complex network model of the boiler. We firstly calculated the Spearman correlation coefficient among different sensor data in a boiler, and then we successfully obtained the correlation coefficient among the sensors. Choosing the boiler steam quantity as the prediction target, we tested the reasonability of the correlation among the sensors we constructed by using SVR model after removing the features with correlation coefficient less than the selected one. If it is reasonable, we constructed a complex network model of the boiler. In general, we use sensors on a boiler as network node, and use the Spearman correlation coefficient among the nodes as the weight of the network sides to construct a complex network model of the boiler. In this model, the sensors are affected by each other. The degree of mutual influence is expressed as the weight of the complex network sides.