Bayesian Networks (BNs) are probabilistic graph models that can be used for classification, prediction, diagnosis, parameter learning and decision making under uncertainty. Probabilities can be inferred from the models and missing values can be imputed, based on probability theory.

BNs are undergoing a renaissance amongst the AI and Machine Learning communities as they provide a fast, efficient probabilistic approach for both supervised (classification) and unsupervised (clustering) problems. Algorithms like Naïve Bayes are often used as the first approximation for classification problems due to its simplicity, yet it often yields results which are comparable to more sophisticated algorithms.

Current research on BNs ranges from the development of new algorithms for learning network structure or Markov Blanket discovery, through to the application of BNs to solve challenging, yet interesting problems in Data Science. Software Libraries are available for parameter learning and the implementation of BNs. Many libraries are available for R and Python (e.g. bnlearn and pomegranate) but Java and C++ libraries are also available. How do these compare against each other? Dynamic Bayesian Networks (DBNs) extend BNs to allow time series or sequences to be modeled. Modelling sequential data is important in many areas of science and engineering and DBNs can be used to examine how probabilities of interest change over time.

The aim of this Special Issue is to gather research and developments relating to algorithms, applications and performance of BNs within the context of Machine Learning and Data Science. Typical topics could include, but are not limited to, the following:

(1) BN structure learning algorithms

(2) Markov Blanket discovery algorithms

(3) Performance analysis of BN models

(4) Comparisons with other Machine Learning algorithms

(5) Evaluation of software tools or libraries

(6) Applications to real-case studies involving either continuous or categorical datasets

(7) Extensions such as Dynamic Bayesian Networks