Learning Bayesian networks: approaches and issues