Although the recommendation system can help users to screen out the desired prod-ucts,and then alleviate the information overload problem.However,most of the current recommenda-tion systems ignore the sparsity of the data,resulting in poor accuracy.Therefore,to improve recom-mendation accuracy,we propose session recommendation models based on graph neural network and temporal coding.The experimental results show that in the 1/4 Yoochoose data set,the accuracy and average reciprocal ranking of the session recommendation results based on graph neural network and temporal coding were 74.3％and 33.9,respectively,which are higher than the other algorithms.More-over,after removing the temporal coding and semantic knowledge base,the accuracy of the top 20 items decreased from 74.5％to 71.6％and 71.1％,respectively,and the average reciprocal ranking decreased from 34.4 to 32.3 and 30.8,respectively.The above results show that the proposed session recommen-dation model performs well,and the time coding and semantic knowledge base can effectively improve the model recommendation accuracy.
维普
万方数据