Research Of Neural Network Prediction Model Based On Modified Whale Optimization Algorithm

Due to the rapid development of intelligent algorithms,our research team is based on the research of increasing population diversity and improving convergence speed and accuracy of the Bat Algorithm(BA).Whale Optimization Algorithm(WOA)for simulating the predation behavior of humpback whales in the ocean,which was proposed in 2016,was carried out in-depth research.Since the research and application of WOA is still in its infancy,the algorithm has the disadvantages of low solution precision,slow convergence speed and easy to fall into local optimum.On this basis,we introduce a global learning mechanism and present a modified whale optimization algorithm based on elite selection strategy(MWOA).Then BP neural network optimization algorithm model based on MWOA is given.Finally,experiments show that the feasibility of MWOA and the strong effectiveness of the optimized neural network algorithm model.The main research work of this paper is as follows:1.Fourteen classic test functions were screened for experimental simulation of BA and WOA algorithms,each of them runs independently 50 times.The simulation proves that regardless of development capabilities,global search capabilities and stability,WOA is more competitive than BA.2.In order to further enhance the global search ability,increase the population diversity of WOA,and avoid it falling into local optimum,we adopt the elite selection strategy and introduce the global learning mechanism.A modified whale optimization algorithm(MWOA)based on elite selection strategy is proposed.The algorithm enhances the global optimization ability of the whale optimization algorithm by means of newly generated whale individuals learning from historically optimal individuals.Then compare the MWOA with BA and WOA through the above 14 classical test functions.Experiments show that the optimization accuracy and convergence performance of MWOA are better than those of BA and WOA.3.In order to capture the advantages of BP neural network,overcome its limitations and improve its convergence speed and global search ability,BA,WOA and MWOA are used to train BP neural networks,respectively.Through four fitting functions and three UCI public data sets,the BP neural network based on MWOA and BP neural network algorithm model based on BA and WOA are compared and analyzed from the aspects of network optimization accuracy and learning performance.Finally,the feasibility and effectiveness of the improved network algorithm model are verified.