COA-SA-BP Algorithm And Its Application In Performance Evaluation Of Energy-saving And Environmental Protection Listed Companies

In recent years,with the increasing complexity of optimization problems in scientific research,evolutionary algorithms have been proposed and improved,but many evolutionary algorithms have the problems of slow convergence speed and low convergence accuracy in the process of finding the optimal solution,and improving the convergence speed and convergence accuracy of evolutionary algorithms in the process of solving optimization problems is a problem that needs to be solved urgently.In addition,the problem of accuracy and efficiency of neural network training methods has been a research hotspot in the field of machine learning.How to optimize the parameters of neural networks and improve the training speed and related performance of neural networks through evolutionary algorithms has become one of the important research directions in the field of machine learning.To address the above problems,this thesis proposes a BP neural network model based on the improved COA-SA algorithm from the following two perspectives,respectively.On the one hand,the SA simulated annealing algorithm,the backward learning strategy,the t-distribution learning strategy and the greedy law are applied to improve the COA coyote optimization algorithm;on the other hand,the parameters of the BP neural network are optimized.The specific approaches are as follows: first,to address the problems of poor global search ability and slow convergence of the COA coyote optimisation algorithm,a COA-SA optimisation algorithm based on the SA simulated annealing algorithm is proposed,and then a dynamic selection strategy with a backward learning strategy and a t-distribution variation operator perturbation strategy is adopted to update the state of the coyote,which improves the performance of the algorithm while avoiding the algorithm from falling into a local optimum.Second,to verify the optimisation capability of the improved COA-SA optimisation algorithm,single-peak and multi-peak functions of different dimensions were selected for optimisation tests to compare the convergence speed and accuracy,stability,running time and other properties of the standard COA algorithm with the improved COA-SA algorithm.The test results show that the improved COA-SA optimization algorithm has better optimization performance compared to the standard COA algorithm.Third,to address the problems of slow training speed and poor fitting of BP neural networks,the improved COA-SA algorithm is applied to optimise the weights and thresholds of standard BP neural networks,and the COA-SA-BP neural network model is proposed.Then,it is compared with the standard BP neural network and COA-BP neural network.Finally,the experimental results show that the generalisation performance and training time of the BP neural network model are effectively improved by this optimisation process.Fourth,the COA-SA-BP neural network model was applied to evaluate the performance of energy-saving and environmental protection listed companies in China in 2020,and the empirical results also showed that the COA-SA-BP neural network model was significantly better than the standard BP neural network model and the COA-BP neural network model in terms of model performance,accuracy and stability.The COA-SA-BP neural network model proposed in this thesis has better correlation performance than the standard BP neural network,which has certain academic and practical significance.