QoS Prediction based on temporal information and request context

Due to the complex and dynamic nature of the Internet, the status of services and their qualities (QoS) change frequently. It is thus important to predict the service quality accurately at runtime from the user’s perspective. Traditional service quality prediction methods either rarely utilize the context data or ignore the request temporal information. As a result, these methods are unable to well capture the depending factors and predict the QoS values accurately. To address this issue, we propose in this paper a novel method, QSPC, to predict service quality concerning both the context data and the temporal information. By mapping raw data to low-dimensional manifold space and fit the real dataset more effectively, our model can greatly utilize the context data to predict the QoS values. Moreover, a sequence-to-sequence layer is proposed to fit the temporal information in the dataset to capture the implicit factors of QoS. The experimental results show that our model outperforms the baseline solutions for service QoS prediction under a benchmark dataset.