A Location-Based Approach for Web Service QoS Prediction via Multivariate Time Series Forecast

With the wide use of service-oriented architecture (SOA) in academia and industry, the method of recommending services based on the Quality-of-Service (QoS) has become increasingly important. To obtain the QoS attributes of Web services, many QoS prediction methods have been proposed. However, existing methods rarely consider the QoS variance caused by the location information and the mutual influence between QoS attributes at the same time. To tackle this problem, this study proposes a novel approach called Mul-TSFL (Multivariate Time Series Forecast based on Location), which combines both collaborative filtering method and time series model to achieve more accurate QoS prediction results. The experiments based on the real-world QoS data set WS-Dream have been conducted to verify the effectiveness of the proposed method. Experimental results show that our method outperforms the related arts.