Prediction model based on neural network and complex plane analysis: a case study of agricultural carbon emissions in Henan Province

Introduction Controlling carbon dioxide emissions and pursuing green development are imperative for global sustainable development. Accurately predicting agricultural carbon emissions is crucial for accelerating emission reduction efforts and guiding green technology innovation. This study focuses on forecasting agricultural carbon emissions in Henan Province to provide data-driven support for green agricultural development. Methods This research utilizes six key influencing factors—chemical fertilizer, pesticide, and agricultural film usage, among others—to predict total carbon emissions. Two primary analytical approaches were employed: a neural network model (comparing Multilayer Perceptron (MLP) and Radial Basis Function (RBF) models) and a nonlinear surface fitting method (specifically, Gaussian multi-modal fitting) for regression and prediction. Results The analysis yielded three main findings: 1) In carbon emission regression, the MLP model demonstrated superior performance with a smaller absolute residual error and significantly higher accuracy (R 2 = 0.998) compared to the RBF model (R 2 = 0.933), establishing it as more suitable for this forecasting task. 2) The Gaussian multi-modal fitting method effectively predicted the time-varying values of the influencing factors (all R 2 > 0.9), enabling reliable further prediction of carbon emissions. 3) Both methods indicate that agricultural carbon emissions in Henan Province follow a quadratic trend over time. The forecast for 2001-2030 reveals a pattern of rapid growth, followed by stable growth, and finally a phase of fluctuating decline. Discussion The high-precision prediction results offer a theoretical reference for advancing green agricultural development in Henan Province. Furthermore, they provide empirical, data-based support for promoting the "green production" concept and disseminating low-carbon policies, thereby enhancing the persuasiveness of ecological education. This contributes to establishing a positive ecological governance cycle of "consciousness - voluntary action - effect translation," ultimately aiding the synergistic enhancement of ecological and social benefits.