一個在不信任的環境下基於服務品質和協同過濾的兩階段網路服務發現機制

Abstract

近年來，隨著網路服務(Web Service)的大量增加，提供相同功能的網路服務也越來越多，以目前Universal Description, Discovery and Integration (UDDI) 所提供以關鍵字為基礎來查詢網路服務功能屬性的方式早已不敷使用，使用者很難從多個相同功能的網路服務中找出一個好的網路服務，甚至所找到的網路服務可能是無法使用的。因此，網路服務的非功能屬性，服務品質(Quality of Service, QoS)就成為一個網路服務查詢的重要指標。然而，以QoS為基礎的查詢方式又衍生出一個重要的問題，因為網路服務的QoS資訊有可能是不正確的，那麼在彼此不信任的環境下，服務提供者所保證的QoS可能會和實際的數值有很大的差距，這樣依據QoS所查詢到的網路服務就會不正確。因此，本論文提出一個以服務品質和協同過濾為基礎的兩階段網路服務發現機制，第一階段先以協同過濾來分析使用者的經驗，過濾QoS不正確的網路服務，第二階段再根據正確的QoS資訊來推薦網路服務。最後，本論文針對所提出的機制實作一雛形系統，並且設計模擬實驗以Average Precision指標來評估本機制的效能，而實驗的結果也指出，本機制的效能可以達96%的精確率，是優於服務品質排序和協同過濾兩個方法。

In recent years, there are an increasing number of Web Services providing similar functionalities. However, at the present time, Universal Description, Discovery and Integration (UDDI) only provide keyword-based service discovery to look up for the functional aspects of the desired Web services. Service consumers are hard to choose a better Web service from those providing similar functionalities. Moreover, the service link that service consumer found probably is unusable. Therefore, Quality of Service (QoS) is becoming an important criterion for selection of the best available service. However, QoS-based service discovery approach has another important issue that the QoS value of Web service may not correct. Consequently, in this untrusted environment, the QoS value service provider guaranteed maybe different from the real QoS value and cause the result of service discovery based on QoS is wrong. Hence, this paper proposes a two-phase Web service discovery mechanism based on QoS and collaborative filtering. In first phase, the mechanism filter out services with incorrect QoS value, and in second phase, the mechanism recommend services with correct QoS value to service consumer. In the end, this paper present simulation experiments to evaluate the effectiveness of our approach using a prototype implementation of the mechanism. The result of experiments also show that the mechanism can reach 96% precision is better than QoS ranking approach and collaborative filtering approach.