A new hybrid logistic regression-naive Bayes classifier

What is it about?

We describe a new classifier that is a hybrid of logistic regression (LR) and naive Bayes (NB), both of which are robust and simple probabilistic classifiers that perform well for a large class of classification problems. We describe how to construct such a classifier with some features in the LR part, and others in the NB part.

Featured Image

Why is it important?

The hybrid LR-NB classifier performs better than pure LR and our NB, and performs as well as state-of-the-art classifiers such as random forest and LASSO. For classification problems with many features, we suggest doing feature selection prior to using our hybrid LR-NB classifier.

Perspectives