An Improved Algorithm for Reducing Bayesian Network Inference complexity

Based on graph search strategy, an improved Bayesian network inference algorithm, which quantitative measure standard is Martelli standard, is presented. The algorithm aims to solve inference complexity. Through proof, the complexity of the improved inference algorithm can reduce from exponential level to polynomial level. In experiment, the algorithm has been applied in the on-line fault diagnosis of motor engineer. The diagnosis result has been compared with that of usual expert system method, and the compared result shows that the improved algorithm can improve the diagnosis efficiency and accuracy. The algorithm has been applied in the on-line fault diagnosis of auto engineer before factory in FAW successfully