Abstract
It is important to develop information systems which are capable of recommending user information or items reflecting user's individual tastes. We propose a music artist recommendation system that provides user with music artists who are the candidates being fit with user's tastes, and are derived from inference based on "favorite music artists" and "reasons to prefer them". Our system consists of an interface for recommendation and a database storing association rules that are patterns of relations among music artists who are liked by people concurrently. The characteristic of our system is to consider individual differences of reasons to prefer music artists in the recommendation processes. The performance of our system shows that reasons to prefer music artists have an effect on ranking recommended music artists, and that our system is superior to other systems in the recommendation accuracies and computational time.
