Web Service QoS Prediction Based on Reputation and Location Aware Matrix Factorization

With the development of cloud computing and Internet technologies, the number of Web services has increased dramatically. It is increasingly difficult for users to locate applicable services among a large number of functionally equivalent candidates. Considering the high cost of time and resources, users cannot invoke all Web services to obtain the desired quality of service (QoS). Therefore, the problem of QoS prediction of Web services has attracted much attention in recent years. Although QoS is often used as a measure of Web service performance, the value of QoS may vary significantly between users depending on their network and geographical location. Furthermore, most traditional approaches perform QoS prediction directly based on historical QoS values provided by users. However, these historical QoS data may contain unreliable values from unreliable users, resulting in significantly lower prediction accuracy. To overcome the above limitations, we propose a reputation and location aware matrix factorization (RLMF) approach for QoS prediction of Web services in this paper. First, we cluster the users and calculate their reputation based on the clustering information through Dirichlet distribution. Then, we integrate the user’s reputation and location information into the matrix factorization model to obtain more accurate prediction results. Additionally, we use Cauchy loss to measure the difference between the observed and predicted QoS values, which makes our approach robust to even outliers. We conducted experiments on a largescale dataset of 1,974,675 QoS values to evaluate our approach. The experimental results show that our approach performs better than state-of-the-art baseline approaches.