Spatio-Temporal Context-Aware QoS Collaborative Prediction

Driven by technologies such as cloud computing,the Internet of Things,and serviceoriented architecture,the service ecosystem is increasingly mature.With the accumulation of service resources,homogeneous services are increasing.How to measure the quality of service from a non-functional perspective,and then recommend suitable service resources for application developers effectively becomes an important application requirement.As a kind of service resource that can be invoked through the Web,Web services are widely used in the Internet production environment.This thesis focuses on the research of quality of service(QoS)collaborative prediction.It mainly focuses on the problem of inadequate consideration and use of the service call context in the prediction process of the existing methods and studies spatio-temporal context-aware collaborative QoS prediction model based on deep learning technology.This thesis proposes different solutions for two different application scenarios.For global prediction task,this thesis proposes a Spatio-Temporal Context-Aware Collaborative QoS Prediction(STCA)model.STCA utilizes the self-attention mechanism to allocate weights to spatial features and fuses them with weights.The interaction between spatial features and temporal feature of service call is captured to predict the QoS values.STCA uses back-propagation methods to optimize model parameters and learn vector representations of temporal and spatial features to achieve the purpose of predicting the quality of service that any user may observe at any time when calling any service.The intensive experiments prove that STCA can significantly reduce the error of QoS prediction,especially in the case of sparse data.As for the task of sequential QoS prediction,this thesis proposes a Spatio-Temporal Context-Aware Sequential QoS Prediction(STCA-S)model.STCA-S utilizes Long Short-term Memory(LSTM)to learn the dependence of temporal features to achieve the purpose of sequential prediction based on STCA.A large number of experiments prove that the accuracy of STCA-S on sequential QoS prediction is significantly improved compared to traditional sequential prediction models.