Bayesian networks with structural restrictions: performance and parallelizations

Bayesian Network algorithms are widely applied in the fields of bioinformatics, document classification, data mining, and marketing informatics.In this paper, three Bayesian Network algorithms are evaluated: Naive Bayes, Tree Augmented Naive Bayes and Ordering-based Bayesian Networks. The algorithms are implemented using Scala, the bnlearn library in R and the WEKA software package. Several data sets with varying levels of attributes are used to test the accuracy of the algorithms and implementation testing is performed across all software platforms. We also parallelize these algorithms for efficiency gains when handling huge data sets. Significant speed-ups were achieved based on parallel processing.