基于显露模式的数据流贝叶斯分类算法

发布时间：2018-03-30 09:54

本文选题：数据流　切入点：显露模式　出处：《软件学报》2017年11期

【摘要】：基于模式的贝叶斯分类模型是解决数据挖掘领域分类问题的一种有效方法.然而,大多数基于模式的贝叶斯分类器只考虑模式在目标类数据集中的支持度,而忽略了模式在对立类数据集合中的支持度.此外,对于高速动态变化的无限数据流环境,在静态数据集下的基于模式的贝叶斯分类器就不能适用.为了解决这些问题,提出了基于显露模式的数据流贝叶斯分类模型EPDS(Bayesian classifier algorithm based on emerging pattern for data stream).该模型使用一个简单的混合森林结构来维护内存中事务的项集,并采用一种快速的模式抽取机制来提高算法速度.EPDS采用半懒惰式学习策略持续更新显露模式,并为待分类事务在每个类下建立局部分类模型.大量实验结果表明,该算法比其他数据流分类模型有较高的准确度.
[Abstract]:However most of the Bayesian classifiers based on schema only consider the support of schema in the target class dataset but ignore the support degree of schema in the set of opposing class data.In addition, for an infinite data flow environment with high speed and dynamic change, a schema-based Bayesian classifier based on static data sets is not applicable.In order to solve these problems, EPDS(Bayesian classifier algorithm based on emerging pattern for data streamer is proposed.This model uses a simple mixed forest structure to maintain the itemsets of in-memory transactions, and adopts a fast pattern extraction mechanism to improve the algorithm speed. EPDS uses semi-lazy learning strategies to continuously update the exposed pattern.A local classification model is established under each class for the transactions to be classified.A large number of experimental results show that the algorithm is more accurate than other data stream classification models.
【作者单位】：北京交通大学计算机与信息技术学院;
【分类号】：TP18;TP311.13

【相似文献】