摘要
朴素贝叶斯分类器的计算过程只有在完全数据库中才成立,而基于相似关系的粗糙集模型具有处理空值的功能,并且提供了属性离散化和约简技术,可以改善属性间的依赖关系。因此,将两种不同的软计算方法相结合,利用粗糙集合理论先把决策表补齐,再对数据进行约简,然后结合朴素贝叶斯分类器,得出分类结果。实验证明这种方法不仅简化了数据和模型的规模,也具有对不完全数据的分类能力。
The naive Bayesian classifier can produce competitive predictive accuracy in many learning tasks, but it can be used only in complete databases. The rough set model based on similarity relationship can process the null, and it has attribute discretization and reduction functions so that the dependency of the condition feature and the decision-making feature can be improved. Therefor, a naive Bayesian classifier algorithm based on the rough set is introduced in this paper. On the basis of the reduction algorithm based on the rough set and the method of processing the null based on the rough set, this method takes into account the influence of the dependency of the condition feature and the decision-making feature on reduction, and gives the most approximate independency reduction results. The experiment result demonstrates that the presented algorithm has perfect performance.
出处
《合肥工业大学学报(自然科学版)》
CAS
CSCD
北大核心
2006年第2期169-172,共4页
Journal of Hefei University of Technology:Natural Science
基金
安徽省教委基金资助项目(2000jl168zd)
关键词
朴素贝叶斯分类
粗糙集合理论
属性约简
native Bayesian classifier
rough set theory
attribute reduction