Structure Learning of Bayesian Belief Networks Using Simulated Annealing Algorithm
Alireza Sadeghi Hesar, Mashhad Branch · 2013
Basically, Bayesian Belief Networks (BBNs) as probabilistic tools provide suitable facilities for modelling process under uncertainty. A BBN applies a Directed Acyclic Graph (DAG) for encoding relations between all variables in state of problem. Finding the beststructure (structure learning) ofthe DAG is a classic NP-Hard problem in BBNs. In recent years, several algorithms are proposed for this task such as Hill Climbing, Greedy Thick Thinning and K2 search. In this paper, we introduced Simulated Annealing algorithm with complete details as new method for BBNs structure learning. Finally, proposed algorithm compared with other structure learning algorithms based on classification accuracy and construction time on valuable databases. Experimental results of research show that the simulated annealing algorithmis the bestalgorithmfrom the point ofconstructiontime but needs to more attention for classification process.