A New-item-first Recommendation Method Based On Time Effect

The rapid development of Internet technology makes it possible to transform contents in various media into sea of bits, and the convenience of releasing information makes people produce information at an unprecedented speed. All of these lead to the overload of information. Information filtering technology becomes very important today. As one way of solving information overload problem, recommendation system has drawn more and more attention and been researched. The traditional recommendation system predicts users' rating in the future basing on the relationship of users and items. As a typical research field, film recommendation has drawn more attention of researchers. However, traditional recommender system has ignored a very import contextual factor when computing the similarity and making recommendations. If the time factor is adopted more effectively, the recommendation quality is not only improved, but also the revenue is increased.Some research is completed in the paper on experimental data which has time line. Main Works are as follows:Firstly, introducing the research background and significance of time-based recommender system, the paper analyzes the status of time-based recommender system. Next, theory and technology of recommender system is introduced.Then a time-based hybrid similarity computing method is proposed in this paper. In order to solve cold-start problem in recommendation, the hybrid recommendation method is proposed, which take PIP similarity method as a prototype.Next, a new recommendation method based on user's “new item preference degree” is proposed in this paper. This method can not only increase the revenue of system by improving the new item recommended ratio, but also alleviate the problem of new item cold start.Finally, some simulation experiments using Movie Lens' film score data set is carried out in this paper, and results of these experiments show that the time-based recommendation strategy proposed in this paper have higher recommendation quality than traditional recommendation method.