Learning Bayesian Network Structure under Incomplete Data Using Maximum Likelihood and Data Augmentation

چکیده مقاله

Bayesian networks are relatively new tools for modeling random processes. Bayesian networks have been used as a categorization in various applications of machine learning, diagnosis of disease, weather forecasting, and forecasting in the stock exchange. These networks are probabilistic graphic models that model relationships between random variables. One of the important issues in Bayesian networks is the process of making them. The purpose of learning the network architecture of Bayesian is to find the best possible structure of the Bayesian network. An important challenge arises when data is incomplete. When the data is incomplete, making the Bayesian network difficult. Maximum resolution is a method for estimating the parameters of a statistical model. When a set of data is performed, a statistical model is obtained, then the maximum likelihood can provide an estimate of the model parameters. After obtaining the data by a data maximization model with a maximization technique, we can obtain the best estimate as the determined data and design the Bayesian network structure. In this paper, a method has been developed combining data Augmentation and maximization Likelihood to build Bayesian networks from incomplete data.

دریافت لینک دانلود مقاله