基于MapReduce的SVM分类算法研究

Research on SVM Classification Algorithm Based on MapReduce

云计算环境中,传统的基于MapReduce的SVM分类算法对数据集的训练是将各子节点训练后得到的支持向量进行合并,得到的分类器分类效率和准确率不理想。为此,文中提出了一种改进的训练算法,在各节点上运用遗传算法来寻找子数据集的最优核函数及参数,用得到的参数组合对子数据集进行训练得到支持向量,合并每个节点训练后的支持向量为全局支持向量,然后在各个节点上将子集与全局支持向量合并作为新的训练数据集。重复这四个步骤,直到全局支持向量不再变化时,则收敛到最优分类模型。最后,经开源云计算平台Hadoop实验验证,该算法的分类正确率比传统的分类算法有了明显提高。

In cloud computing environment,the method adopted by the traditional SVM sorting algorithms based on MapReduce of train-ing data set is too simple and it just merges support vectors after nodes’ training,so the efficiency and accuracy of classifier are not very ideal. To solve the problem above,an improved training algorithm is proposed in this paper. Firstly,use the genetic algorithm to get the optimal kernel function and parameters on each node at the same time,then using the combination to train the data set for support vector, and afterwards,combining all support vectors after training as a global support vector,and then merging every data subset with global support vector on each node to get a new training data set. Repeat these four steps until the global support vector no longer changes and that’ s to say,it converges to the optimal classification model. Finally,the experiment on Hadoop proves that the classification accuracy of new algorithm is improved obviously than traditional classification algorithms.