Correlation-Based Top-k Recommendation for Web Services

The growth and popularity of Web services make personalized recommendation as one of the most important research issues in service computing, which can help service users find the optimal Web services from large functionally equivalent service candidates according to the Quality-of-Service (QoS). Although significant works have been done on improving prediction accuracy by employing collaborative filtering (CF) approaches, some of the well-known approaches ignore relationships among Web services in neighbor selection of recommendation which have an important impact on QoS prediction accuracy. Different from these works which simply rank the neighbors according to their similarity values, we propose a correlation-based top-k recommendation approach for Web services that leverages the correlation graph and corresponding Web service ranking matrix. Our approach aims to improve the process of finding k neighbors by using correlation among Web services instead of the similarity. Experiments are conducted by employing real-world Web service QoS data to evaluate the prediction accuracy of our approach. The experimental results demonstrate that our approach achieves better QoS prediction results than other competing approaches.