Real data often contains a mixture of discrete and continuous variables, butmany Bayesian network structure learning and inference algorithms assume allrandom variables are discrete. Continuous variables are often discretized, butthe choice of discretization policy has significant impact on the accuracy,speed, and interpretability of the resulting models. This paper introduces aprincipled Bayesian discretization method for continuous variables in Bayesiannetworks with quadratic complexity instead of the cubic complexity of otherstandard techniques. Empirical demonstrations show that the proposed method issuperior to the state of the art. In addition, this paper shows how toincorporate existing methods into the structure learning process to discretizeall continuous variables and simultaneously learn Bayesian network structures.
展开▼
