Improving the K2 Algorithm Using Association Rules Parameters

A Bayesian network is an appropriate tool to work with a sort of uncertainty and probability, that are typical of real-life applications. Bayesian network arcs represent statistical dependence between different variables. In the data mining field, association rules can be interpreted as well as expressing statistical dependence relations. K2 is a well-known algorithm which is able to learn Bayesian network. In this paper we want to present an extension of K2 called K2-rules that exploits a parameter normally defined in relation to association rules for learning Bayesian networks. The experiments performed show that K2-rules improves K2 with respect to both the quality of the learned network and the execution time