Information Technology Journal1812-56381812-5646Asian Network for Scientific Information10.3923/itj.2006.540.545HengXing-Chen QinZheng WangXian-Hui ShaoLi-Ping 3200653A new approach to learning Bayesian networks (Bns) was proposed in this study. This approach was based on Particle Swarm Optimization (PSO). We start by giving a fitness function to evaluate possible structure of BN. Next, the definition and encoding of the basic mathematical elements of PSO were given and the basic operations of PSO was designed which provides guarantee of convergence. Next, full samples for the training set and test set are generated from a known original Bayesian network with probabilistic logic sampling. After that, the structure of BN was learned from complete training set using improved PSO algorithm steps. Finally, the simulation experimental results also demonstrated sthis new approach’s efficiency.]]>Eberhart, R.C. and J. Kennedy,1995Heckerman, D., D. Geiger and D.M. Chickering,1995Henrion, M.,1988Lauritzen, S.L. and D.J. Spiegelhalter,1988