Learning Structure of Bayesian Network Using Ant Colony Algorithm Assisted by Genetic Algorithm

Abstract

Ant colony algorithm (ACA) has been applied on structure learning for Bayesian Network since it is accurate to solve optimization problem, but its speed is slow at initiation phase. This paper proposes an ACA based structure learning approach improved by genetic algorithm (GA), which is fast in initiation phase. Let GA learn the structure of Bayesian Network from training data quickly, and then take the rough outcome produced by GA to initiate ACA in both pheromone matrix and states of ants, finally the structure is worked out accurately .Through a series of tests, this approach is proved to be accurate and fast compared to traditional ways.