Study of Optimizing Web Service Composition – Using Genetic Algorithm and Case-base Clustering

• Main Authors: Kang-Wei Peng, 彭康韋
• Other Authors: Hong-Wen Chen
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/33p8hg

碩士 === 銘傳大學 === 資訊傳播工程學系碩士班 === 102 === In recent years, more web services are provided by enterprises as popular applications over Internet. With the growth of web services, a single service cannot usually match the demands of users; and thus, the necessity of combining services is rising and named as Web Services Composition. Since the quality and restrictions of web services become more unstable to handle as the increasing number of services, the issues of finding most suitable web services for users is more important today. In the study, the items of web services were searched through the server based on users’ functional service requirements. For the quality factors of web services, subjective weights of each individual user were evaluated by the technique of analytical hierarchy process. If the number of web services to meet the user’s demands found through searching is small, the best composition of web services would be suggested by applying dynamic programming to take the ensemble quality of integrated services into consideration. On the other hand, if the number of available web services is too large, it becomes intractable for dynamic programming to process. The proposed restricted genetic algorithm, which was initialized with the optimal service composition recorded within similar cases, would be used to find the best or near-best composition of web services by simultaneously considering the personal weights of service quality for a user and the specified composition constraints of each individual web service. The technique of clustering is also applied in the case database to assist us to efficiently select similar cases out. In addition, some experimental evaluations were designed to validate the effectiveness and efficiency of the proposed method. Thus, the proposed method is highly expected to improve the management and applications of clouding web services.