基于C-Vine Copula理论的监督学习分类器的优化

Optimization for C-Vine Copula-Based Supervised Learning Classification

由于朴素贝叶斯分类器对特征变量作了独立性假设,忽略了相关性,导致在某些特征相关的情况下分类效果很差。为了提高分类效果,本文对有缺失的数据集利用C-Vine Copula理论进行填补从而得到完整的数据集,并结合Copula函数研究特征变量之间的相关性优化问题,用C-Vine Copula分类器对完整数据集做分类。结果表明,基于C-Vine Copula理论的监督学习分类器具备良好的分类性能。

Because of the feature independence assumption, the correlation between variables is ignored, causing that the Naive Bayes works poorly in classification for some cases when the features are correlated. In this paper, for improving the classification effect, the missing datasets are filled by using C-Vine Copula theory. As a result, the complete datasets are got after imputation. By combining the copula function and investigating on the correlation between features, C-vine copula classifier is used to classify complete datasets. The obtained results show that the supervised learning classifier based on the C-Vine Copula theory has better performance.