Exploiting hierarchical domain values for Bayesian learning

Authors:
Yiqiu Han;Wai Lam
Affiliations:
Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong, Shatin, Hong Kong;Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong, Shatin, Hong Kong
Venue:
PAKDD'03 Proceedings of the 7th Pacific-Asia conference on Advances in knowledge discovery and data mining
Year:
2003

Citing 7
Cited 0

On the Optimality of the Simple Bayesian Classifier under Zero-One Loss

Machine Learning - Special issue on learning with probabilistic representations
Distributional clustering of words for text classification

Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval
Hierarchical classification of Web content

SIGIR '00 Proceedings of the 23rd annual international ACM SIGIR conference on Research and development in information retrieval
Probabilistic hierarchical clustering for biological data

Proceedings of the sixth annual international conference on Computational biology
Multidimensional Data Integration and Relationship Inference

IEEE Intelligent Systems
Finding an Optimal Gain-Ratio Subset-Split Test for a Set-Valued Attribute in Decision Tree Induction

ICML '02 Proceedings of the Nineteenth International Conference on Machine Learning
Improving Text Classification by Shrinkage in a Hierarchy of Classes

ICML '98 Proceedings of the Fifteenth International Conference on Machine Learning

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper proposes a framework for exploiting hierarchical structures of feature domain values in order to improve classification performance under Bayesian learning framework. Inspired by the statistical technique called shrinkage, we investigate the variances in the estimation of parameters for Bayesian learning. We develop two algorithms by maintaining a balance between precision and robustness to improve the estimation. We have evaluated our methods using two real-world data sets, namely, a weather data set and a yeast gene data set. The results demonstrate that our models benefit from exploring the hierarchical structures.