Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
426008 | Future Generation Computer Systems | 2012 | 10 Pages |
Many of the current Quality of Service (QoS) based web service selection systems assume that users will formulate QoS queries accurately. However, this may not always be the case. It would be helpful if users could browse through the QoS data of the service repository before they start the searching process. In this way, they could know what the actual QoS value distributions are, so that they could put down reasonable numbers in their QoS queries. The browsing process performed on QoS queries, on the other hand, could help service providers understand the actual QoS demands from users. In this paper, we propose to use the Scatter/Gather model—a well-known browsing model for large document collections, to implement QoS browsing. We modified the original model to make it fit QoS data and implemented a few variations of the model. We tested and compared performance on both synthetic and real QoS datasets, mainly focusing on the ability to handle real time interaction with users while balancing efficiency and accuracy.
► QoS browsing is important for understanding QoS value distribution of web services. ► Scatter/Gather model can be used to implement QoS browsing. ► Symbolic clustering algorithm is the main component of our QoS browsing system. ► Improved LAIR2 implementation of Scatter/Gather model achieves both high accuracy and efficiency.