微型电脑应用

针对只关注用户历史信息的网络信息推荐模型无法准确为用户推荐适当信息的问题，从用户的纵向历史信息维度与横向社交信息维度共同入手，在纵向历史信息维度上设计以门控循环单元(GRU)和图神经网络模型为主的用户物品模型，在横向社交信息维度上设计以多头注意力机制为主的用户社交模型，进而形成同时兼顾用户横纵向信息的网络信息推荐模型。并利用性能对比试验与消融试验来检验模型的应用效果。试验结果显示，学习率为0.0005,GRU层数为2层时，多头注意力机制和门控循环单元模型的性能是最佳的。同时与同类型模型相比，所设计的MR模型在Epinions数据集与Ciao数据集环境下的均方根误差与平均绝对误差都更小。由此可见，MR模型能够在网络信息推荐时进行更加全面的综合性信息分析，进而实现更加准确的信息推荐。

In view of the problem that the network information recommendation model that only focuses on the user's historical information, and cannot accurately recommend appropriate information for users, considers the vertical historical information dimension and the horizontal social information dimension of users. In the vertical historical information dimension, a user item model based on the gated recurrent unit(GRU) diagram neural network model is designed, and in the horizontal social information dimension, a user social model based on the multi-head attention mechanism is designed, thus, forming a network information recommendation model that takes into account both the horizontal and vertical information of users. The performance contrast test and ablation test are used to test the application effect of the model. The research results show that the performance of the multi-head attention mechanism fusion gated recurrent units model is the best when the learning rate is 0.0005 and the number of GRU layers is 2. At the same time, compared with the same type of model, the root mean square error and mean absolute error of the MR model designed in the study are smaller in the environment of Epinions dataset and Ciao dataset. It can be seen that the research and design of MR model can carry out more comprehensive information analysis when recommending network information, and then achieve more accurate information recommendation.