针对分布环境的数据挖掘要求,提出了基于支持向量机的分布数据挖掘模型DSVM.定义了DSVM中特征多叉树的概念,描述了基于移动Agent访问分布数据集来构建特征多叉树的方法,阐述了通过特征多叉树来反映分布环境各数据集属性总体特征的思想,并利用该数据结构和支持向量机的特点,提出了基于壳向量的分布式支持向量机增量算法来修正和完善特征多叉树,最终实现分布环境下全局的数据挖掘.实验结果表明,该模型有效地解决原有分布环境下其他挖掘算法存储开销大、执行效率差、安全性和隐私性低等问题.

The paper presented a distributed data mining model based on Support Vector Machines DSVM. It described the definition of multi-branches tree of Eigen (ET) and the method of building ET based on mobile Agents accessing to distributed datasets. It elaborated the concept by using ET to reflect the characteristic of attribute in the distributed dataset, and then proposed the algorithm of distributed incremental Support Vector Machines based on hull vector (HDIS) using the data structure of ET and the feature of Support Vector Machine. Finally, the ET was modified and improved by HDIS to realize distributed data mining. The experimental results show the DSVM providing high capability and efficiency of distributed business data mining.