A new collaborative filtering recommendation approach based on naive Bayesian method

Abstract

Recommendation is a popular and hot problem in e-commerce. Recommendation systems are realized in many ways such as content-based recommendation, collaborative filtering recommendation, and hybrid approach recommendation. In this article, a new collaborative filtering recommendation algorithm based on naive Bayesian method is proposed. Unlike original naive Bayesian method, the new algorithm can be applied to instances where conditional independence assumption is not obeyed strictly. According to our experiment, the new recommendation algorithm has a better performance than many existing algorithms including the popular k-NN algorithm used by Amazon.com especially at long length recommendation. © 2011 Springer-Verlag.