A deep learning approach for collaborative prediction of Web service QoS

Abstract

Web services are a cornerstone of many crucial domains, including cloud computing and the Internet of Things. In this context, QoS prediction for Web services is a highly important and challenging issue that facilitates the building of value-added processes such as compositions and workflows of services. Current QoS prediction approaches, like collaborative filtering methods, mainly suffer from obstacles related to data sparsity and the cold-start problem. Moreover, previous studies have not conducted in-depth explorations of the impact of the geographical characteristics of services/users and QoS ratings on the prediction problem. To address these difficulties, we propose a deep-learning-based approach for QoS prediction. The main idea consists of combining a matrix factorization model based on a deep autoencoder (DAE) with a clustering technique based on geographical characteristics to improve the effectiveness of prediction. The overall method proceeds as follows: first, we cluster the input QoS data using a self-organizing map that incorporates knowledge of geographical neighborhoods; by doing so, we can reduce the data sparsity while preserving the topology of the input data. In addition, the clustering step effectively overcomes the cold-start problem. Next, we train a DAE that minimizes the squared loss between the ground-truth QoS and the predicted one, for each cluster. Finally, the unknown QoS of a new service is predicted using the trained DAE related to the closest cluster. To evaluate the effectiveness and robustness of our approach, we conducted a comprehensive set of experiments based on a real-world Web service QoS dataset. The experimental results show that our method achieves better prediction performance than several state-of-the-art methods.