Selecting suitable web services based on the quality-of-service (QoS) is essential for developing high-quality service-oriented applications. A critical step in this direction is acquiring accurate, personalised QoS values of web services. As the number of web services is enormous and the QoS data are highly sparse, improving the accuracy of QoS prediction has become a challenging issue recently. In this study, we propose a novel QoS prediction model, called neural fusion matrix factorisation, wherein we combine neural networks and matrix factorisation to perform non-linear collaborative filtering for latent feature vectors of users and services. Moreover, we consider context bias and employ multi-task learning to reduce prediction error and improve the predicted performance. Furthermore, we conducted extensive experiments in a large-scale real-world QoS dataset, and the experimental results verify the effectiveness of our proposed method.

根据服务质量 (QoS) 选择合适的 Web 服务对于开发高质量的面向服务的应用程序至关重要。朝着这个方向迈出的关键一步是获取准确、个性化的 Web 服务 QoS 值。由于 Web 服务数量庞大且 QoS 数据高度稀疏，提高 QoS 预测的准确性成为最近的一个具有挑战性的问题。在这项研究中，我们提出了一种新的 QoS 预测模型，称为神经融合矩阵分解，其中我们结合神经网络和矩阵分解对用户和服务的潜在特征向量执行非线性协同过滤。此外，我们考虑了上下文偏差并采用多任务学习来减少预测误差并提高预测性能。此外，