[Show abstract][Hide abstract] ABSTRACT: In this paper, we focus on how to overcome several limitations in the traditional research of collaborative filtering(CF).
We present a novel CF recommendation algorithm, named DSNP(Dynamic Similar Neighbor Probability). This algorithm improves
the neighbors’ similarities computations of both users and items to choose the neighbors dynamically as the recommendation
sets. How to select the confident subsets which are the most effective neighbors to the target object, it is the first stage.
A major innovation is by defining a dynamic neighbor probability over the trustworthy subsets. Moreover, we define a prediction
algorithm that combines the advantages of dynamic neighbor coefficient with the user-based CF and the item-based CF algorithms.
Experimental results show that the algorithm can achieve consistently better prediction accuracy than traditional CF algorithms,
and effectively leverage the result between user-based CF and item-based CF. Furthermore, the algorithm can alleviate the
dataset sparsity problem.