A Max-relevance-min-divergence Criterion for Data Discretization with Applications on Naive Bayes

MRmD is a supervised discretization framework that can balance discrimination power and generalization ability.

Implementation

This is a source code of MRmD for a naive Bayes classifier using MATLAB. The version of MATLAB should be >=2019b.

• To successfully run the code, you need to install the Bioinformatics Toolbox and Statistics and Machine Learning Toolbox.

• Download the source file, place it in MATLAB and then run the main.m file.

Dataset

We provide a sample dataset flare.data2 for one of the used datasets flare-solar in the paper. The detailed descriptions of all datasets used in the paper can be found at https://archive.ics.uci.edu/ml/index.php.

If you use this code, please cite:

@article{wang2024max,
  title={A max-relevance-min-divergence criterion for data discretization with applications on Naive Bayes},
  author={Wang, Shihe and Ren, Jianfeng and Bai, Ruibin and Yao, Yuan and Jiang, Xudong},
  journal={Pattern Recognition},
  volume={149},
  pages={110236},
  year={2024},
  publisher={Elsevier}
}