摘要
通过对支持向量数据描述SVDD(Support Vector Data Description)算法的阐述和SVDD算法在增量学习过程中支持向量集变化特性的分析,提出一种新的SVDD增量学习算法。以Spambase邮件语料库作为实验数据源,将其与非增量学习算法以及一般传统增量学习算法进行比较,结果证明,该算法在保证垃圾邮件识别精度的同时又大大缩短了训练时间。
This paper presents a novel approach for Support Vector Data Description (SVDD) incremental learning algorithm through the elaboration of SVDD and the analysis of support vector set' s change rule in incremental learning process. With the aid of spambase email cor- pus as the experimental data source, the comparison was carried on with non-incremental learning algorithm as well as general traditional learn- ing algorithm. The experimental result indicates that this approach guarantees the spam recognition precision and reduces the training time greatly as well.
出处
《计算机应用与软件》
CSCD
2009年第9期237-239,共3页
Computer Applications and Software