Prediction of Agricultural Carbon Emission Mechanism Based on Improved Neural Network Search Algorithm

Based on the improved search algorithm, this paper optimizes the BP neural network model, selects three variables, namely planting GDP, aquaculture GDP and fishery GDP, as predictors to predict farm agricultural carbon emissions, and uses five-fold cross-validation to evaluate the prediction results. It overcomes the shortcomings of the poor stability of the traditional BP neural network, and all evaluation indexes are better than the model optimized by a single algorithm, which proves that the improved sparrow search algorithm greatly improves the global search ability and optimization accuracy of the neural network, and has an obvious effect on the performance optimization of the neural network. The improved search algorithm optimizes the BP neural network model, which is a prediction model of agricultural carbon emissions with strong stability, accuracy and robustness. The results show that the aquaculture industry plays a dominant role in the impact of total carbon emissions, and controlling the scale of aquaculture can effectively control the total carbon emissions.