Adaptive learning algorithms for Bayesian network classifiers

This thesis is concerned with adaptive learning algorithms for Bayesian network classifiers (BNCs) in a prequential (on-line) learning scenario capable of handling the cost-performance trade-off and concept drift. All these algorithms are integrated into the adaptive prequential framework for supervised learning, AdPreqFr4SL. We evaluated our algorithms on artificial domains and benchmark problems and show their advantages and future applicability in real-world on-line learning systems. © 2008 - IOS Press and the authors. All rights reserved.