Using New Data to Refine a Bayesian Network
This addresses a specific gap in Bayesian network refinement for researchers and practitioners, but it is incremental as it builds on prior work.
The paper tackles the problem of refining an existing Bayesian network's structure using new data that may only cover a subset of variables, developing a new approach based on the Minimal Description Length principle and an adapted learning algorithm, with experimental evidence showing its effectiveness.
We explore the issue of refining an existent Bayesian network structure using new data which might mention only a subset of the variables. Most previous works have only considered the refinement of the network's conditional probability parameters, and have not addressed the issue of refining the network's structure. We develop a new approach for refining the network's structure. Our approach is based on the Minimal Description Length (MDL) principle, and it employs an adapted version of a Bayesian 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 effectiveness of our approach.