Recommendation Diversity Improvement Method Based on Neural Factorization Machine
The neural factorization machine model solves the problem of click-through-rate prediction in the sce-nario of sparse data,but only focusing on the prediction accuracy leads to the poor recommendation diversity effect of this model.Aiming at this problem,this paper proposes a diversity enhancement method based on neural factorization machine.This method constructs a diversity input matrix through the interaction history of users and items,and uses the user activity and item popularity to modify the diversity input matrix.The diversity input matrix is integrated into the neural factorization machine model in different schemes,which is used as an additional information source to en-hance the diversity expression ability of the neural factorization machine model.The experimental results show that on three datasets with different degrees of sparsity,MovieLens,Film Trust and Book-Crossing,the proposed method can greatly improve the recommendation diversity with a small loss of recommendation accuracy.
NETL
NSTL
万方数据
