Abstract:

The item-based collaborative filtering algorithm (IBCF),a recommendation algorithm with high precision,simple and easy to use in actual system, is widely used in the field of recommendation systems. But it meets a higher computational time complexity for similar calculation because of the long length of item vector. In this paper, a sampled approach firstly is suggested to represent an item vector called mean model item vector representation through analyzing theory of IBCF algorithm, to solve the problem of the long length of item vector and cut down the computational time. Experiments using Movie Lens datasets show that the algorithm is very efficient to cut down the computational time on the premise of accuracy. Furthermore, some right sampling methods can be used to optimize the calculation method of similarity in order to meet practical application requirement.

Key words: similarity computing; mean model; recommendation system; itembased collaborative filtering