A Theoretical and Experimental Evaluation of Augmented Bayesian Classifiers
نویسندگان
چکیده
Naive Bayes is a simple Bayesian network classifier with strong independence assumptions among features. This classifier despite its strong independence assumptions, often performs well in practice. It is believed that relaxing the independence assumptions of naive Bayes may improve the performance of the resulting structure. Augmented Bayesian Classifiers relax the independence assumptions of naive Bayes by adding augmenting arcs that obey certain structural restrictions, between features of a naive Bayes classifier. In this paper we present algorithms for learning Augmented Bayesian Classifiers with respect to the Minimum Description Length (MDL) and Bayesian-Dirichlet (BD) metrics. Experimental results on the performance of these algorithms on various datasets selected from the UCI Machine Learning Repository are presented. Finally, a comparison of the learning rates and accuracies of various Augmented Bayesian Classifiers is presented on artificial data.
منابع مشابه
Comparing Bayesian Network Classifiers
In this paper, we empirically evaluate algorithms for learning four Bayesian network (BN) classifiers: Naïve-Bayes, tree augmented Naïve-Bayes (TANs), BN augmented NaïveBayes (BANs) and general BNs (GBNs), where the GBNs and BANs are learned using two variants of a conditional independence based BN-learning algorithm. Experimental results show the GBNs and BANs learned using the proposing learn...
متن کاملAdjusting Dependence Relations for Semi-Lazy TAN Classifiers
The naive Bayesian classifier is a simple and effective classification method, which assumes a Bayesian network in which each attribute has the class label as its only one parent. But this assumption is not obviously hold in many real world domains. Tree-Augmented Naive Bayes (TAN) is a state-of-the-art extension of the naive Bayes, which can express partial dependence relations among attribute...
متن کاملMulti-dimensional Bayesian Network Classifiers
We introduce the family of multi-dimensional Bayesian network classifiers. These classifiers include one or more class variables and multiple feature variables, which need not be modelled as being dependent on every class variable. Our family of multi-dimensional classifiers includes as special cases the well-known naive Bayesian and tree-augmented classifiers, yet offers better modelling capab...
متن کاملStatistical Classifiers in Computer Vision
This paper introduces a unified Bayesian approach to 3–D computer vision using segmented image features. The theoretical part summarizes the basic requirements of statistical object recognition systems. Non–standard types of models are introduced using parametric probability density functions, which allow the implementation of Bayesian classifiers for object recognition purposes. The importance...
متن کاملEvaluation of the Performance of the Markov Blanket Bayesian Classifier Algorithm
The Markov Blanket Bayesian Classifier is a recentlyproposed algorithm for construction of probabilistic classifiers. This paper presents an empirical comparison of the MBBC algorithm with three other Bayesian classifiers: Naïve Bayes, Tree-Augmented Naïve Bayes and a general Bayesian network. All of these are implemented using the K2 framework of Cooper and Herskovits. The classifiers are comp...
متن کامل