Research on the E-commerce Knowledge Graph Link Prediction Problem Based on the Relation Scale Model
To address the issues of low accuracy in link prediction models for e-commerce knowledge graphs and the repeated recommendation of the same type of products,an improved Relation Scale(RS)model was proposed.The strength of relationships between the head and tail entities of triples was assessed using the TransE and TuckER model.The weights of all relationship paths were determined by introducing a scaling factor,thus enhancing the model's convergence speed.Experimental results show that the MRR,Hits@1,Hits@3 and Hits@10 of the improved models are all enhanced based on the OpenBG500 dataset.The MRR and Hits@10 for the RSTransE model increase by 47.4%and 71.1%respectively,compared with the traditional TransE model.The MRR and Hits@10 for the RSTuckER model increase by 35.8%and 28.4%respectively,compared with the traditional TuckER model.These findings indicate that the RS model can more accurately predict user needs and achieve more personalized and precise recommenda-tion results.
国家科技期刊平台
NSTL
万方数据
