QoS Attributes Prediction with Attention-based LSTM Network for Mobile Services

With the rapid popularization of mobile terminal equipment, the traditional service is developing towards the direction of mobile service composition with high real-time and high reliability. As mobile services and applications continue to be more widespread, it is necessary to accurately predict user-side Quality of Service (QoS), to ensure the optimal selection, composition and adaptation of mobile services. The time-varying characteristic of QoS is the commonness of web services. In addition, user's position may change at intervals in mobile service scenario. And QoS varies from place to place due to various external factors. Many existing QoS prediction approaches fail to consider the mobility of users and cannot reflect the correlation between time varying QoS values and some external factors such as user's position. To solve this problem, we use a LSTM based Encoder-Decoder network with attention mechanism to predict the QoS at future time. This model combines the location variability of service users and the fluctuation of QoS in time series. Experiments are conducted on WS-DREAM dataset and results have proved the effectiveness of the proposed method.