论文部分内容阅读
大数据背景下,如何对海量数据进行挖掘是目前研究的一个热点问题。序列最小最优化(SMO)算法是实现支持向量机(SVM)对大数据挖掘的有效算法。现有算法假定核函数是正定或半正定,限制了核函数的选择。为解决这一不足,提出了针对非半正定核v-SVR的SMO算法。所提算法不仅适用于非半正定核,而且具有较好的回归精度。
Under the background of big data, how to mine massive data is a hot issue in current research. The sequence minimum optimization (SMO) algorithm is an effective algorithm for support vector machine (SVM) for big data mining. Existing algorithms assume that the kernel function is positive definite or semi-positive definite, which limits the choice of kernel function. To solve this problem, an SMO algorithm for non-positive definite v-SVR is proposed. The proposed algorithm not only applies to non-positive definite nuclear, but also has a good regression accuracy.