Bayesian network learning based on relationship prediction PSO and its application in agricultural expert system

To resolve the problem for modeling agricultural expert system effectively in the complicated and uncertain agricultural production system, a Bayesian network learning algorithm based on relationship prediction Particle Swarm Optimization (PSO) is proposed. A successful interpretation of data goes through discovering crucial relationships among variables, and such a task can be accomplished by a Bayesian network. However, when lots of variables are involved, the learning of the network slows down and may lead to wrong results. In this study, we demonstrate the feasibility of applying an existing Particle Swarm Optimization (PSO)-based approach with mutual information for filtering the irrelevant attributes of the dataset, resulting in candidate Bayesian networks which provide the optimization direction for BN learning and searching. Experimental tests carried out with both artificial data and real data coming from the agricultural domain. Experimental results demonstrate that the presented algorithm is effective and efficient, which can be used in the agricultural expert system.