摘要
传统数据挖掘方法,主要针对静态数据进行挖掘,而对数据流挖掘往往失效。为了解决数据流的数据挖掘问题,提出一种通过改变传统支持向量机增量式学习方法,利用轮转式结构将多分类器按照数据流时间顺序进行组合,并且通过对分类器的优化,可以提高模型对数据流分类的准确率并减少训练时间消耗。实验结果表明,该模型在保证学习精度和推广能力的同时,提高了训练速度,适合于数据流在线分类和在线学的问题。
The traditional data mining method mainly is used to analyze the static data, but it is invalidation for data stream. In order to solut the problem of data stream data mining, through modifying the traditional SVM based incremental leafing, employing the cycle structure to update the multi-classifier and the classifier ordered by time. Apart from this, the optimization is applied for improving the classification accuracy and reducing time consuming. The final experimental result shows the proposed approache of SVM-based incremental learning is considerably faster than the standard SVM and the classical incremental algorithm.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第11期4225-4229,共5页
Computer Engineering and Design
关键词
增量式学习
支持向量机
网络异常检测
概念漂移
多分类器模型
incremental learning
support vector machine
network intrusion detection
concept drift
multipleclassifier models
