ULMF: Web Service QoS Collaborative Prediction with Explicit Ratings and Implicit User Location

Abstract

Since more and more Web services with equivalent function but different QoS are available in Internet, predicting unknown QoS value is often required for Web service selection and composition. Previous prediction approaches underestimate the role of user location information, which have a significant impact on user QoS experience according to our empirical analysis on public real-world QoS dataset-WSDream. In this paper, we proposed a personalized Web service QoS collaborative prediction method, which extends matrix factorization model by smoothly incorporating both explicit QoS values user rated in the past and implicit user location information that inherently existed in rating-oriented model. Experimental results show that compared with other approaches, suggested method in this paper can achieve higher prediction accuracy and as well as performs well in cold start situation.