Abstract

Reliability prediction is an important way to evaluate the performance of Web Service, which would help Service Matching to achieve the available services to the target users. Current reliability prediction approaches commonly use the collaborative way to provide the personalized predictions for different users. However, these approaches lack the consideration of the time-effectiveness of the past reliability records and treat the past reliability records collected at any time equally. Also, these approaches are unable to predict for the target users who have insufficient past reliability records. To address the problems, we propose an improved time-effectiveness reliability prediction approach of Web Service, which employs a hybrid time-effectiveness model to evaluate the time-effectiveness of the past reliability records, and utilizes the users' geographical data to find the nearby users who would aid the prediction. On the real reliability datasets of Web Service, we discuss the effects of various settings of the parameters, and achieve the proper settings of the parameters. Comparative experiments unfold that our approach manifests the lower predication errors than other approaches.

Paper Details