The Optimization To Hybrid Recommendation Algorithms In Video Website

ABSTRACT:The advent of the Internet age makes the rapid development of the online video industry and the appearance of many excellent video websites. In order to meet the needs of different users, the number of users and items of modern video websites increase exponentially, and this has made the user-item matrix become more and more sparse which of course will encounter cold starting problem when conducting personalized recommendation. This situation, in combination with somewhat inappropriate similarity calculation methods currently used, will surely reduce the quality of recommender system gradually. Hybrid recommendation models and methods, which combine collaborative and content-based approaches, alleviate the cold starting problem to some extent, but the results are still unsatisfactory.To promote the recommendation quality and provide the users with real personalized service, we present an optimized recommender algorithm which is based on a hybrid model. In our algorithm, the similarity function is linear combination of the item property similarity and an optimized modified cosine similarity. The weighting factor, which is generated automatically, is related to the number of users who have rated both items. In the item attribute similarity computation, the weighting of each attribute is included, and TF-IDF algorithm is used. The modification to the cosine similarity measure considers both the rating tendency and activity from the users. To deal with the cold starting problem, we also acquire user similarity through user property information with weighting factors computed by SVDFeature. As experimental results demonstrated, our algorithm effectively improves the recommendation quality of the video websites whereas alleviate cold starting problem resulting from both users and items.