Dynamic food web prediction and optimization based on deep learning and genetic algorithm

This study proposes a dynamic food web prediction and optimization model based on deep learning and genetic algorithm, aiming to improve the prediction accuracy and stability analysis ability of dynamic changes in agricultural ecosystems. The model integrates the long short-term memory network (LSTM-Attention) with attention mechanism to effectively capture the nonlinear dynamic characteristics of species number and pesticide concentration, and optimizes key parameters through NSGA-II genetic algorithm to achieve stable balance of ecosystem. Experimental results show that the model is significantly better than traditional methods in prediction accuracy, with the mean square error reduced to 1.82×10^-3, the determination coefficient reaching 0.96, and the maximum eigenvalue real part reaching -0.35, which verifies its application value in sustainable agricultural management. This study provides efficient tools and theoretical support for the scientific modeling and management of agricultural ecosystems.