Dynamic analysis and application of data-driven green behavior propagation on heterogeneous networks

Abstract

Clear waters and lush mountains constitute invaluable assets, and the sustainable development of the energy economy relies on green behavior. This paper establishes a Centrist–Positive–Negative system for the propagation of green behavior on heterogeneous networks by considering the transition mechanisms among individuals with different attitudes. The equilibrium points of the system are computed, and the sufficient and necessary conditions for positive equilibrium points are provided. We analyze the necessary conditions for Turing instability and the first-order conditions for parameter identification based on optimal control. Numerical simulation results indicate that various network structures can influence the timing of Turing bifurcation. Moreover, the presence of heterogeneity within networks exacerbates the instability of solutions. Media publicity and government management notably exert an inverted U-shaped influence on outcomes. Furthermore, the homogeneity or heterogeneity of the networks should not affect the effectiveness of parameter identification. Utilizing accurate data from the Policy Research Center for Environment and Economy and the China National Environmental Monitoring Centre, we conduct parameter identification on the effectiveness of government management in 13 cities in Jiangsu Province in 2021, yielding promising results. Upon comparison of three time series forecasting models, the LSTM model demonstrates superior performance. A parameter identifying the effectiveness of government management through the prediction of comprehensive air quality indices by using LSTM neural networks yields similarly favorable outcomes. Extending the network to a larger scale further enhances identification performance.