Cognitive Graphs for Socio-Economic Objects Management

The article is devoted to the study of methods to improve the use of cognitive graphs in automated analytics and the management of socioeconomic objects and processes based on the impact on carbon dioxide emissions of following the ‘green agenda’ in different countries (the European Union, Southeast Asia, etc.). The broad context of ongoing research in different countries is complemented by the analysis and modeling of carbon dioxide emission factors, for which graph neural networks were used. The application, construction and main components of graph neural networks (GNN) and the implementation of the connection prediction algorithm are discussed in detail. Machine learning was carried out using an ensemble machine learning method to determine the contribution of each feature to the classification or regression process. The random forest was chosen as a regressor to combine bagging and random subspace techniques. An extensive set of decision trees was used, which individually are not capable of providing the proper quality of estimates, but thanks to consolidation, a satisfactory result is achieved. The results of the study confirmed the initial hypothesis that the main influence on global carbon dioxide emissions is exerted by countries with developed industrial production, which do not consistently adhere to the ‘green agenda’. According to the model, of 44 countries, only 15 have an impact on the world in terms of carbon dioxide emissions. In turn, the greatest influence on the world is exerted by countries such as China, the USA, Iran, Russia, India, and Japan.