[2D6-GS-2] Machine learning: Bayesian estimation

[2D6-GS-2-01] Learning Bayesian network classifiers using Integer Programing

〇Maomi Ueno¹, Kentaro Inamura¹, Kouya Kato¹, Shouta Sugahara¹ (1. The University of Electro-Communications)

Keywords:Bayesian network, Classifier, Bayes

Bayesian network classifier is known to have high explainability with maintaining high prediction accuracies as same as that of Deep learning. The unique feature is to assymptotically estimate the true classification probability. However, earlier learning method requires two stages process and entailes expencive memory complexity. To resolve this difficulty, this study propsoes a new learning Bayesian network classifiers method using the integer programing method. Results demonstrate the proposed method improves the accuracies and possible learning network size.