Research Of Ensemble Extreme Learning Machine Based On Hybrid Rice Optimization Algorithm

As a kind of single hidden layer feed-forward neural network,the extreme learning machine(ELM)has been widely used due to its high training speed.However,the initial randomization of extreme learning machine parameters leads to unstable classification results.In order to improve the extreme learning machine classification accuracy,enhance robustness,and generalization,the evolutionary algorithm is usually used to optimize its parameters.But most of the optimization process is easy to fall into a local optimum and the convergence speed is slow.The hybrid rice optimization algorithm is a newly proposed natural calculation method,which has the characteristics of high optimization speed and optimization ability.In this paper,the hybrid rice optimization algorithm was used to optimize parameters and optimize ensemble extreme learning machines.The main contents are as follows:1)The basic principles of hybrid rice optimization algorithm and extreme learning machine are briefly introduced.Using hybrid rice optimization algorithm to optimize the parameters of extreme learning machine.By testing the extreme learning machine model after optimizing the parameters on the common data set,and then comparing the results with other commonly used optimization algorithms.The fitness of the function is worthwhile.The hybrid rice optimization algorithm has a good ability for searching the best result.At the same time,by comparing the optimized extreme learning machine model with other commonly used classifiers on Weka,the classification accuracy indicates that the optimized extreme learning machine classifier model has better classification ability.2)In order to further study the extreme learning machine classifier model,this paper integrates the basic limit learning machine classifiers with different excitation functions.In the ensemble process,each base classifier is given different weights,and the weight value will affect the classification result.Therefore,the hybrid rice optimization algorithm is used again to optimize the base classifier weights.The optimized integrated extreme learning machine model was tested on the common data set,and the results were compared with other optimized models.The results show that the hybrid rice algorithm has a better search capability and the ensemble extreme learning machine's performance is better than others.In general,this paper focuses on the study of the extreme learning machine and the hybrid rice algorithm.By optimizing the parameters of the extreme learning machine,a single extreme learning machine classifier with optimal parameters is obtained.By further optimizing the weights of the ensemble extreme learning machine,improving the generalization performance and robust performance of the classifier model.The experimental results show that the hybrid rice optimization algorithm has better searching ability,and the ensemble extreme learning machine model optimized by hybrid rice algorithm has higher classification accuracy and generalization performance.There are broad application prospects in the field of classification problems.