A Bayesian network is a representation of a joint probability distribution of a set of random variables with a possible mutual causal relationship. Bayesian networks aim to model conditional dependence, and therefore causation, by representing conditional dependence by edges in a directed graph.