No abstract

No abstract

The product recommendation research has been focusing on modelling users' reviews to construct the relation of users and products. Thus, the recommended performance can be improved by obtaining virtual ratings from corresponding reviews. However, these perspectives on reviews do not take into account the product field characteristic, which may impact the recommendation performance. To this point, this paper proposes a hybrid collaborative filtering approach to compute the correlation value considering product attributes. First, Product Attribute Weight and Product Attribute Score are introduced to formalize the product attributes for user and product respectively in a quantitative way. After that, the recommended ranking formula for the new model is presented. Finally, we carry out experimental analysis to show our method can effectively improve the performance of recommendation under a sparseness dataset.