BAYESIAN NETWORK STRUCTURAL LEARNING FROM DATA: AN ALGORITHMS COMPARISON

摘要

The manual determination of Bayesian Network structure or, more in general, of the probabilistic models, in particular in the case of remarkable dimensions domains, can be complex, time consuming and imprecise. Therefore, in the last years the interest of the scientific community in learning bayesian network structure from data is considerably increased. In fact, many techniques or disciplines, as data mining, text categorization, ontology description, can take advantages from this type of processes. In this paper we will describe some possible approaches to the structural learning of bayesian networks and introduce in detail some algorithms deriving from these ones. We will aim to compare results obtained using the main algorithms on databases normally used in literature. With this aim, we have selected and implemented five algorithms more used in literature. We will estimate the algorithms performances both considering the network topological reconstruction both the correct orientation of the obtained arcs.