Summary
Bayesian Networks (BNs) are graphical models which represent multivariate joint probability distributions which have been used successfully in several studies in many application areas. BN learning algorithms can be remarkably effective in many problems. The search space for a BN induction, however, has an exponential dimension. Therefore, finding the BN structure that better represents the dependencies among the variables is known to be a NP problem. This work proposes and discusses a hybrid Bayes/Genetic collaboration (VOGAC-MarkovPC) designed to induce Conditional Independence Bayesian Classifiers from data. The main contribution is the use of MarkovPC algorithm in order to reduce the computational complexity of a Genetic Algorithm (GA) designed to explore the Variable Orderings (VOs) in order to optimize the induced classifiers. Experiments performed in a number of datasets revealed that VOGAC-MarkovPC required less than 25% of the time demanded by VOGAC-PC on average. In addition, when concerning the classification accuracy, VOGAC-MarkovPC performed as well as VOGAC-PC did.
See the full content of this document
Extract
An Optimized Evolutionary Conditional Independence Bayesian Classifier Induction Process
1. Introduction-section
Bayesian Networks (BNs) are graphical models which represent multivariate joint probability distributions described by directed acyclic graphs (DAGs). These models have been used successfully in several studies in many application areas. The techniques that have been developed are still evolving, but many researches dealing with this theme revealed that the BN learning algorithms can be remarkably effective in many problems. BNs can be used in unsupervised as well as in supervised learning tasks. When a supervised learning task is conducted, the BN is usually called Bayesian Classifier (BC) and this nomenclature is applied also in this work.The good results achieved in the last years motivated the development of many BNs learning algorithms [11]. However, the search space for a BN with n variables induction has an exponential dimension. Therefore, finding the BN structure that better represents the dependencies among the variables is known to be a NP problem [4], thus it is hard to identify the best solution for all the application problems [1]. In this sense, approximate models can be induced and the search space of this process is likely to be reduced. To do so, some restrictions are usually imposed and often the algorithms obtain good results with acceptable computational effort.There are basi...See the full content of this document
Sponsored links
