【摘要】 为满足用户对Web服务的高质量要求，提出一种基于图卷积的Web服务QoS预测算法，结合多任务学习使算法可以在一次预测中得到两种属性的QoS值。挖掘用户和服务之间隐藏的图结构信息，将用户和服务的编号信息转换成嵌入向量，得到初始特征，经过图卷积与特征融合处理操作获得更深层的特征信息，经过降噪自编码器对特征进行重构，提高算法的鲁棒性。在公开的QoS调用数据集上进行实验，并与3种方法对比，实验结果表明，该算法在各项评价指标上总体表现更好，为图神经网络在QoS预测方向的研究提供了思路。更多还原

【Abstract】 To meet the high quality requirements of users for Web services, a graph convolution-based QoS prediction algorithm for Web services was proposed and combined with multi-task learning, so that the algorithm could obtain the QoS values of two attributes in a prediction. The hidden graph structure information between users and services was mined, the numbering information of users and services was converted into embedding vectors to obtain the initial features, the deeper feature information was obtained through graph convolution and feature fusion processing operations, and the features were reconstructed by noise reduction self-encoder to improve the robustness of the algorithm. Experiments were conducted on the publicly available QoS call dataset and compared with the three methods. Experimental results show that the algorithm performs better in all evaluation indexes, which provides ideas for the research of graph neural network in the direction of QoS prediction.更多还原