Predicting QoS for Cloud Services through Prefilling-Based Matrix Factorization

Quality of service (QoS) is an important indicator that users need to focus on when choosing services from the cloud center to build a service-based system. However, it is often difficult for users to collect the QoS records of all services under consideration in real circumstances. In this paper, we proposed an adjusted matrix factorization (MF) model named PFMF to predict the missing QoS values so as to make a scientific decision on service selection. Different from the existing MF models, in our PFMF method, the training is performed on the prefilled QoS matrix rather than the original sparse matrix. The prefilling of matrix is implemented by improving the PCC-based CF method. The key improvement is to consider the fluctuation degree of QoS records when finding the neighbor users or services. To validate the prediction performance of the proposed method, the comparison experiments are conducted on a widely-used large-scale dataset. The experimental results show that our PFMF method outperforms both the MF-based methods and PCC-based CF methods. In addition, it also shows good stability.