Explanation Guidance of Augmentation Multi Pair Contrastive Learning Sequential Recommendation
Comparative learning in sequence recommendation can effectively alleviate the problem of data sparsity,but the existing random enhancement methods used in comparative learning produce positive samples that ex-hibit false positives.Single pair comparative learning cannot effectively neutralize the problem of false negatives in negative samples,resulting in limited recommendation performance.To address the above issues,an explanation guided enhanced multi pair contrastive learning sequence recommendation algorithm(EMC4Rec)is proposed.Firstly,the explanation method is used to determine the importance scores of the items in the user sequence,and then the Ex-planation Guidance Augmentation is used to generate multiple high-quality positive samples according to the impor-tance scores of the items and perform multi-pair contrastive learning.Experimental results on Beauty,Toys and ML-1 datasets show that EMC4Rec'stop-k Hit Ratio(HR)and Normalized Discounted Cumulative Gain(NDCG)are 8.13%and 11.38%higher than those of CL4SRec and CoSeRec respectively.
万方数据
维普
