Multiple similarity collaborative filtering recommendation among users

Abstract [Objective] Through the analysis of multiple similarity among users, the problem that the traditional user based collaborative filtering algorithm only uses a single similarity and leads to the decline of recommendation quality is solved. [Method] The original single similarity calculation formula is improved, and the multiple similarity calculation formula is put forward, on this basis, the multiple similarity prediction score is calculated. [Result] By comparison with the traditional user based collaborative filtering algorithm, the method put forward in this paper has outstanding effect. [Limited] Users’ interests will change with time, so time information should be included in the calculation. [Conclusion] From the experiment, we can find that the improved method has better recommendation quality than traditional methods.