Abstract:

The major objecive of recommendation algorithms is to accurately predict users’rating values of the unknown product items.However,it has recognized that an accurate prediction of rating values is not the only way for achieving user satisfaction,and the diversity of recommendation lists has gained importance recently.In this study,we propose a novel method which can adjust the weights of product categories to enhance the diversity levels of their own recommendation lists with little decrease in accuracy.Experimental results show that our method outperforms other methods,which has a very low computational time complexity and is highly scalable.

Key words: recommendation system;diversity;recall