A Hybrid Genetic Algorithm for Bayesian Network Optimization

To find an optimized structure in a Bayesian network is a NP problem. How to get a network with a high score is an important question. In this paper, we discuss some theories about Bayesian network study and propose a hybrid genetic algorithm HGA-BN for Bayesian network optimization. The algorithm is based on genetic algorithm, uses simulated annealing technology to select its children, and uses self-adaptive probabilities of crossover and mutation to do local search. When the computation converges, we use hill-climbing algorithm to optimize the result, which can enhance the ability of local search.