Temporal-aware QoS Prediction based on Tensor Factorization and Self-Attention for Cloud Services
Wenyu Tang, Mingdong Tang, Fenfang Xie · 2023
The widespread adoption of cloud computing has given rise to numerous cloud services with various functionalities, which are very useful in creating cloud applications. To ensure high reliability of cloud applications, it is crucial to invoke the right cloud services with optimal quality of service (QoS) in the runtime. Thus, predicting QoS of cloud services dynamically becomes a need. Previous methods do not take into a full consideration of the relationships between users, services, and time, thus their models’ expressive ability is limited. Based on the intricate relationships between users, services and time, this paper proposes a temporal-aware QoS prediction approach via exploiting tensor factorization and self-attention representation (TFSA) for cloud services. TFSA firsty utilizes tensor factorization to historical QoS data and leverages a self-attention mechanism to refine the personalized representations of users, services and time. Then, TFSA integrates the original and personalized representations for QoS predictions. Extensive experiments conducted on a real-world dataset show that the proposed approach significantly outperforms the other state-of-the-art methods with better prediction accuracy.