Learning Parameters for Hybrid Bayesian Network

Abstract

The Hybrid Bayesian network (HBN) is a type of Bayesian network (BN) with continuous and discrete variables. A continuous random variable can have both discrete and continuous parents, whereas a discrete random variable can have only a discrete random variable. Bayesian network (BN) is a graphical representation of the joint distribution of the random variables. BN is used to identify the structure in a large amount of data and captures the conditional independencies among random variables, thereby used for prediction of the unseen or missing data. In real-world scenarios, most of the data is a combination of both discrete and continuous distributions. In this work, parameters are estimated for a continuous random variable in HBN, which is a linear combination of both continuous and discrete parents, also estimating joint multivariate Gaussian distribution for continuous random variable having only continuous parents.