A Bayesian Network Based Structure Learning Algorithm
Long Zhang · 2016
To solve the difficulties of high calculation quantity and low precision in constructing an air transport network, the paper puts forward a kind of hybrid algorithm, which called Bayesian network model based conditional independence test and heuristic search (CIBNS, CI-based Bayesian Network Search). The new algorithm firstly makes use of the conditional independence test to compress the search space, which to ensure the solution quality and speed up the search process simultaneously. Then, the method introduces the heuristic search based on the BDeu Measure score to improve the efficiency of constructing. Experimental results on simulated and real data show that the new algorithm can effectively construct aviation network. Its performance in terms of efficiency and accuracy is better than the hill climbing method and local search method. The solution quality is better and the convergence speed is faster. The CIBNS achieves a better balance in terms of validity and the calculated efficiency of the solution.