Supervise Learning With Copulas
Xiaoping Shen, Robert L. Ewing, Jia Li · 2019
The naïve Bayes classifier plays an important role among the classifiers based on supervised learning, although it requires strong condition on the feature independence assumptions. A measurement for the independency checking in the data preprocessing is necessary to guarantee the effectiveness of the classifier. Copula Theory is a mathematical tool in dependency modeling. In this paper, we recall elements of copulas and introduce a new algorithm to construct multiscale copula estimators which can be used for the independency testing to improve the accuracy of the Naïve Bayes classifier.