Learning Bayesian Network Structures Based on MDL and Hybrid Genetic Algorithms
Ronggui Wang · Microelectronics & Computer · 2002
One of the difficulties of the application of Bayesian Networks is that when the data arise, it is very hard to learn the structures of Bayesian Networks from large databases. Both Standard Genetic Algorithms and Hill-Climbing can be used in structure learning, but none of them can get proper result easily. The combination of the two algorithms can have better effect. The ALARM Network is learned, Hybrid Genetic Algorithms is usedthen Bayesian Network structure is got as the result after the Minimum Description Length is selected as the fitness function.