Research On Distributed Hybrid Rice Algorithm Based On Hadoop

Hybrid Rice Optimization Algorithm(HRO)is a new heuristic evolutionary computational swarm intelligence optimization algorithm inspired by the breeding methods of hybrid rice.The three lines are CMS,maintainer,and restorer.The CMS lines are crossed with the maintainer lines to produce hybrid rice.Restoration maintains good attributes through self-intersection.The process of hybridization is an evolutionary process.Self-interaction process is the process of group search.These two processes affect each other and are combined in appropriate proportions.The algorithm has fewer parameters,a simple principle,easy implementation,and strong versatility.Compared with other algorithms,it has good stability,strong searching ability,low computational complexity,fast calculation speed,and is applicable to various optimization problems.However,with the rapid development of Internet information technology,the information data generated directly or indirectly also grows exponentially,which poses new challenges to traditional data mining algorithms.Under such circumstances,the concept of big data cloud computing has emerged.Researchers have merged traditional data mining algorithms with emerging technologies,used distributed computing capabilities to improve the performance of the algorithm,and achieved good results.The Hadoop distributed computing framework has become the main solution for big data processing.Based on the previous work and experience,this paper designs and implements a distributed hybrid rice algorithm based on Hadoop,aiming to solve the problem of improving the performance of HRO algorithm in the case of large-scale data volume.Through the optimization of the kernel parameters of the kernel function of the support vector machine,it is further verified that the Hadoop-based distributed hybrid rice algorithm has a good effect on the parameter optimization.The main work of this article is as follows:I.Through researching and learning the theoretical models of the hybrid rice optimization algorithm,such as algorithm models and implementation principles,and consulting a large number of domestic and foreign literature on the solution of distributed group intelligent optimization algorithms,a distributed hybrid rice algorithm based on Hadoop was designed and implemented.There are two kinds of distributed solutions.Through the experimental verification,the best solution is chosen to conduct in-depth research on the algorithm.Comparing the Hadoop-based distributed hybrid rice algorithm and the stand-alone hybrid rice algorithm by setting the size of the population.Through the increase of the number of Hadoop cluster nodes,the above two algorithms are compared to each other.The experimental results show that under certain other conditions,when the population size increases,the distributed hybrid rice algorithm based on Hadoop performs better than the single hybrid rice algorithm,and with the increase of the number of Hadoop cluster nodes,its advantages become more and more obvious.II.This paper studies the parameter optimization problem of support vector machines,and proposes a method to optimize the nuclear parameters of SVM kernel function based on Hadoop's distributed hybrid rice algorithm.Since mature hybrid group intelligent optimization algorithms such as hybrid rice optimization algorithm,genetic algorithm,particle swarm optimization algorithm,and ant colony algorithm all belong to the probability algorithm,when designing experiments,the average value obtained through a large number of experiments is used for comparison.This is more illustrative of the problem.In the optimization of SVM parameters,the experimental comparison of Hadoop-based distributed HRO-SVM and stand-alone HRO-SVM algorithm on running time and classification accuracy was conducted through the change of hybrid rice population.The experimental results show that the Hadoop-based distributed HRO-SVM algorithm is basically the same as the standalone HRO-SVM algorithm with the increase of the population,but it is obviously superior to the traditional standalone HRO-SVM in running time.