面向会话推荐的融合改进GNN的人工智能统计学方法研究

Research on Artificial Intelligence Statistical Method for Session Recommendation Based on Fusion Improved GNN

为提升会话推荐模型的推荐精度,研究通过构建入度与出度矩阵对图神经网络进行改进,并基于此设计出一种智能会话推荐模型。结果表明,在全部数据集中此次设计的模型P@20与MRR@20数值分别为71.4与19.1,高于对比模型,且它在不同会话子序列下的P@20与平均倒数排名@20的标准差分别为0.26与0.35。测试数据证明,此次设计的会话智能推荐模型推荐精度较传统方法更高,在智能客服、智能交互系统中具有一定应用潜力。

In order to improve the recommendation accuracy of the session recommendation model,this study improved the graph neural network by constructing an in and out matrix,and designed an intelligent session recommendation model based on this.The test results are as follows:In all datasets,the model P@20 designed in this design is different from MRR@20 The values are 71.4 and 19.1 respectively,which are higher than the comparison model,and its standard deviation between P@20 and the average reciprocal ranking@20 under different session subsequences is 0.26 and 0.35,respectively.Test data shows that the session intelligent recommendation model designed this time has higher recommendation accuracy than traditional methods,and has certain application potential in intelligent customer service and intelligent interaction systems.