Adaptive and Collaborative Recommendation using Content Type
Ki-Tae Han, Moon-Kyoung Park, Yong Suk Choi · Jeongbo gwahaghoe nonmunji. so'peuteuweeo mich eung'yong · 2011
This paper proposes an adaptive and collaborative recommendation method using content type, which can improve performance considerably by alleviating sparse matrix and cold start problems. If conventional methods don't have user's rating data when a new user comes into the system or don't have enough data for some users, they can't recommend any content to the users due to insufficiency of matrix data. To resolve these problems, we propose a user-content_type matrix with relatively higher density than conventional user-content matrix. Using user-content_type matrix, we compute user's preference for a content type and then reflect it to the prediction of preference for each content. A method of reflection is combining prediction of preference for a content with prediction for its type. Using a MAE(Mean Absolute Error) and Coverage measures for performance evaluation. We identify significant performance improvement compared to existing collaborative recommendation methods.