Documents Categorization Based on Bayesian Spanning Tree
Hui-Feng Shi, Tiegang Fan, Guoli Zhang · 2006
In this paper, an algorithm of learning a simple type of Bayesian network-Bayesian spanning tree with maximum log-likelihood is presented. The log likelihood function is used to measure the Bayesian spanning tree with respect to given documents data. In a Bayesian spanning tree, besides the root node, each node has at most two parent nodes. The Bayesian spanning tree is an unsupervised classifier called Bayesian spanning tree classifier. Under the Bayesian spanning tree classifier, documents in documents set are categorized. The experimental result indicates that Bayesian spanning tree classifier is more effective and has higher accuracy