Using New Data to Re ne a Bayesian Network

We explore the issue of re ning an existent Bayesian network structure using new data which might mention only a subset of the variables. Most previous works have only considered the re nement of the network's conditional probability parameters, and have not addressed the issue of re ning the network's structure. We develop a new approach for re ning the network's structure. Our approach is based on the Minimal Description Length (MDL) principle, and it employs an adapted version of aBayesian network learning algorithm developed in our previous work. One of the adaptations required is to modify the previous algorithm to account for the structure of the existent network. The learning algorithm generates a partial network structure which can then be used to improve the existent network. We also present experimental evidence demonstrating the e ectiveness of our approach.