Location-Based Web Service QoS Prediction via Preference Propagation for Improving Cold Start Problem

With the popularity of service-oriented architecture, many web systems have been developed in form of composite services. Since the performance of these composite services highly depends on Quality of Service (QoS) of employed atomic web services, it is important to predict the QoS values of atomic web services with high accuracy. Although collaborative filtering based approaches have recently been proposed to predict the web service QoS values, they mostly face a cold start problem which causes unreliable prediction due to the highly sparse historical data, newly introduced users and web services. Furthermore, existing work only considers the case of newly introduced users. In this paper, we propose a Location-based Matrix Factorization technique via Preference Propagation (LMF-PP) to improve the cold start problem in web service QoS prediction domain. LMF-PP exploits the location information of entities (i.e., Users and web services) and employs the preference propagation to make the accurate QoS prediction even for the newly introduced entities and in the small amount of data (i.e., Highly sparse matrix). The performance of LMF-PP is compared with that of existing approaches on a real world dataset. The experimental results show that LMF-PP can outperform the existing approaches in not only a cold start environment but also a warm start environment.