Learning Bayesian Networks using expert’s prior information on structures

Most of the approaches developed in the literature to elicit the a priori distribution on Directed Acyclic Graphs (DAGs) require a full specification of graphs. Nevertheless, expert's prior knowledge about conditional independence relations may be weak, making the elicitation task troublesome. This paper presents and evaluates an elicitation procedure for DAGs which exploits prior knowledge on network topology. The elicitation is suited to large Bayesian Networks (BNs) and it accounts for immediate causal link and DAG sparsity. We develop a new quasi-Bayesian score function, the P- metric, to perform structural learning following a score-and-search approach. We tested our score function on two different benchmark BNs by varying sample size and prior belief concerning structures. Our results show the effectiveness of the proposed method and suggest that the use of prior information improves the structural learning process.