一种基于RippleNet模型的推荐精度提高方法

A Method for Improving Recommendation Accuracy Based on RippleNet Model

RippleNet模型引入向量表示的同时充分利用实体连接关系,挖掘高阶语义,实现精准推荐,但并没有充分考虑到数据的重要性。通过构建概念图谱的最大子网,消除数据的冗余,提高RippleNet模型的推荐精度。利用构建最大子网的思想,通过最大子网以消除原始数据冗余。处理冗余数据后,对比原始数据,在Top-k场景中不同k值的平均准确率提高1%,在CTR点击率预测场景下所得到的平均AUC值从91.3%提高到91.9%。实验表明,通过提取最大子网可以提高推荐精度。

The RippleNet model introduces vector representation while making full use of entity connection relationships,mining high-level semantics,and achieving accurate recommendation but does not fully consider the importance of data.This paper eliminates data redundancy and improves the recommendation accuracy of the RippleNet model by constructing the largest subnet of the concept map.Using the idea of constructing the largest subnet,the original data redundancy is eliminated by constructing the largest subnet.After processing redundant data,the average accuracy of different k values in the Top-k scenario compared to the original data increased by 1%;the average AUC value obtained in the CTR click-through rate prediction scenario increased from 91.3%to 91.9%.Experiments show that the recommendation accuracy can be improved by extracting the largest subnet.