An improved optimization algorithm for Bayesian network structure based on scaling framework
Finding the optimal network structure is an NP-hard problem for Bayesian networks'probabilistic inference.In order to accurately model the causal relationships between nodes,this paper proposes a learning algorithm with an improved network structure based on a scaling framework.Firstly,the scaling framework is used for causal analysis to determine the strength of causal relationships between nodes through the slope matrix.This result is used as the basis for constructing the network search space,and the initial score of the network structure can be improved.Secondly,the coati optimization algorithm based on scoring methods is used to find the network structure with the highest score.Thus,the scoring search ability in Bayesian networks is enhanced.Finally,the structure with the highest score is processed by the add-arc,the subtract-arc and the steering-arc operations,and the optimal structure with the highest degree of fitting is found.Simulation experiments are conducted on standard networks with different complexities,and the results show that the proposed algorithm converges faster,can find the optimal structure in a shorter time,and has a higher score of structure learning and a higher convergence accuracy.These indicate that the algorithm has more advantages in accuracy and search efficiency.
国家科技期刊平台
NSTL
万方数据
