Mass customization in internet marketing and the overwhelming amount of information to be processed each day lead internet users to a growing interest in collaborative recommendation systems which suggest products and services to users. Collaborative Filtering (CF) has been used successfully at e-commerce sites like Amazon.com and JCPenny.
Despite its popularity, CF has some limitations. In traditional CF, all the ratings are weighted equal regardless of the error they contain. This is problematic, because each rating can have different level of error.
In this paper, an adaptive collaborative filtering algorithm is proposed. The suggested algorithm captures the level of error that each rating contains, and utilizes the captured knowledge to increase the adaptability of the collaborative filtering recommendation. The concept of the adaptive weight proposed to express the relative informational value of each rating in numerical form. The performance of suggested algorithm is verified by conducting an experiment.