To find an optimal elimination ordering for Bayesian networks, a multi-heuristic-based ant colony system named MHC-HS-ACS is proposed. MHC-HS-ACS uses a set of heuristics to guide the ants to search solutions. The heuristic set can evolve with the searching procedure in an adaptive way.
MHC-HS-ACS also utilizes a heuristic-based local search to accelerate its convergence. Computational experiments show that MHC-HS-ACS can find very high quality solutions.