Algorithms For Bayesian Network Structure Learning Based On The Greedy Search

Bayesian network is a network model indicating dependence and independence relationships between random variabiles, and is also called bays net, causal probability network or belief network. Combining probability theory and graph theory, Bayesian network provides a natural and intuitional method discovering potential relationships from data. There has been a great deal of research focused on Bayesian network in machine learning, artificial intelligence, and probability inference due to their uncertain knowledge expression, the ability to match the probability and property for learning based on prior knowledge. At present, algorithms for constructing Bayesian network structures from data have become one of research hotspots.Firstly, the research background knowledge and basic theory of Bayesian network are presented and application of the greedy search in structure learning is studied. On the basis of this theory, an improved algorithm constructing the essential graph based on the adjacency matrix of network structure and a new algorithm constructing structure of Bayesian network from data are presented in order to overcome shortcomings of the greedy search. These are the foundation for discussing the hybrid algorithm.Secondly, advantages and disadvantages of the immune algorithm are analyzed and, on the basis of immune algorithm, the hybrid algorithm for learning Bayesian network is described combining with advantages of the greedy algorithm. Selection of initial framework and fitness function in addition to the simplified conditional independence test are mainly discussed in order to optimize the hybrid algorithm. Numerical experiments show that more optimal structure can be established according to this algorithm and the goals for building Bayesian network structures can be realized.Finally, summary is depicted and directions of future research, such as learning algorithm of missing data, structure learning of dynamic Bayesian network, intelligent algorithm applied in Bayesian network and application of Bayesian network in biological engineering, are pointed out.