Hill climbing algorithm for Bayesian network structure

Abstract

Bayesian Network (BN) model is a method developed to describe causal relationships between variables in a system. BN is a form of Probabilistic Graphical Model or a simple probabilistic graph built from Bayes probability theory and graph theory. There are two approaches used to construct the BN structure, namely the constraint-based method and the score-based method. The scored-based method is a method that assigns a score to each structure in BN and tries to maximize the scoring with several heuristic search algorithms. The scoring is executed through the usage of Bayesian Information Criterion (BIC) scoring function. In this study, scored-based totally is solved through the Hill Climbing (HC) algorithm. This algorithm is a value-based algorithm in a directed graph space and includes a heuristic search method that works greedily. The results of the study show that the use of the score-based method followed by the HC algorithm and the BIC scoring function can build the BN structure.