One Bayesian Network Construction Algorithm Based On Dimensionality Reduction

Bayesian network is a common probabilistic graphical model. It often can achieve good result in expression of uncertainty for knowledge and regular as well as in data classification. Since Bayesian network construction is a relatively complex problem, we propose a Bayesian network construction dimensionality reduction algorithm (BNDR). This algorithm maps a set ofassociated features to an abstract feature node by feature clustering and mapping. It can ensure no loss in accuracy and improve the time efficiency. For more complex industrial scenario, using the BNDR, you can get better practical efficiency.