CAJ | 학술논문

为了提高数据的分类效率和准确度，利用云计算提供的弹性集群平台来解决计算力伸缩性瓶颈，并用MapReduce编程模型对SVM进行Map和Reduce并行化处理，并将基于优化理论的遗传算法（ GA）引入SVM分类算法中对分类器参数进行优化，以分类器的准确率作为GA算法适应度函数，找出全局最优的模型参数和核函数参数值。经开源云计算平台Hadoop实验验证，数据分类的准确度有了明显的提高，整个分类过程的加速度几近呈线性增加。
위료제고수거적분류효솔화준학도，이용운계산제공적탄성집군평태래해결계산력신축성병경，병용MapReduce편정모형대SVM진행Map화Reduce병행화처리，병장기우우화이론적유전산법（ GA）인입SVM분류산법중대분류기삼수진행우화，이분류기적준학솔작위GA산법괄응도함수，조출전국최우적모형삼수화핵함수삼수치。경개원운계산평태Hadoop실험험증，수거분류적준학도유료명현적제고，정개분류과정적가속도궤근정선성증가。
By Support Vector Machines ( SVM) classification of large-scale data on traditional platform, plat-form computing power usually encounters with scalability bottlenecks and classification process cannot be effi-ciently parallelized.That leads to low classification efficiency.The values of the SVM model parameters and kernel function parameters decisively affect the classification accuracy .But the values of those parameters most-ly are random values or the values of experience, which results in low classification accuracy.In order to im-prove the accuracy and efficiency of data classification , the elastic cloud computing cluster platform is used to solve the scalability bottleneck of computing power , so that SVM can map and reduce parallelization by MapRe-duce programming model.Genetic algorithm (GA) based on optimization theory is introduced to optimize the parameters of the classifier in SVM classification algorithm , with accuracy of the classifier as the fitness function of GA algorithm to find the global optimum parameter values of SVM model and kernel functions .Experimental results show that in the open source cloud computing platform Hadoop , data classification accuracy is signifi-cantly improved and the acceleration of the entire classification process is almost a linear increase .