An online personalized recommendation model based on Bayesian networks

Abstract

It is one of an important method of using Bayesian networks in electronic commercial recommended system. But the models of Bayesian networks for describing recommended system have a problem that it could not learn online. The paper puts forward an online personalized recommended model based on Bayesian networks. The paper uses a partial ordering to represent previous structure and find posterior distributions of every node on the orders to realize online structure learning. It also uses a correctional function to revise log likelihood for online parameter learning. The experiment shows that the model can be learned online for personalized recommended system. © 2008 International Federation for Information Processing.