Research on Crop Prediction Optimization Based on BP Neural Network and Particle Swarm Algorithm

Abstract

Accurate crop yield prediction is crucial for optimizing agricultural production and ensuring food security. However, traditional methods are limited by their simplicity and inability to cope with complex environments and climate changes, leading to insufficient prediction accuracy. This study integrates crop planting data from rural areas of a website for the year 2023, employing a BP Neural Network-PSO Hybrid Model. By conducting correlation analysis on existing crop data from cultivated land and combining the maximization of profit and minimization of costs, the optimal Crop Planting Strategy is determined. In the prediction process, the sigmoid function is cautiously selected as the activation function, and key parameters required for pre-adjustment in machine learning are set to ensure the model's stability and convergence speed. The training target error is set to 0.00001 to pursue higher fitting accuracy. After validation, the MSE Mean Squared Error Value is obtained as 0.085, which sufficiently indicates the model's high rationality and prediction accuracy. When importing data for prediction, the maximum total profit over seven years is calculated to be 4,135,919,090.20 CNY. By comparing the fit degrees of the values before and after, it is clear to observe an upward trend in crops. Finally, Sensitivity Analysis is used to test the stability of the model after achieving the optimal solution.