Bi-lateral Mapping And Evolution Of Requirement Patterns And Service Patterns

It is still a great challenge to make coposite services that fulfill massive requirements from large community of customers efficiently. Service network(SN) approach has been put forward to deal with this issue in a cost-effective and agile way, i.e., a large number of services are connected as a customizalbe SN in terms of underlying semantics correclations, and when a new requirement arrives, the SN is customized to generate a composite service solution. However, SN still generates a composite solution from the very beginning. Consider a group of similar requirements, it is a valuable research to further improve the efficiency of SN customizaiton process.Based on existing research and theory of SN and SN customization, this paper propose a new bi-lateral patterns based service network customization algorithm, and then design a method to evaluate the importance of elements in SN, finally, give some analysis of the evolution of bi-lateral patterns and related elements in SN.1. The total amout of customer requirement is huge so that the efficiency of personalized customization is relatively low whereas the cost is still high. However, it has been observed that there are similarities among requirements, and so are there among composite solutions of these requirements. In this paper, we identify bi-lateral patterns(i.e., requirement patterns and service patterns) from historical service composition records, and then establish the probabilistic mappings between them; consiquently, a bi-lateral patterns based service network customization algorithm named BPSC is put forward to take full advantage of such priori knowledge to speed up the customization process.2. Among service composite solutions, elements in SN are used together to fulfill a specific requirement, but due to the characteristics of a group of historical requirements, the usage and the importance of elements may show differences. Allocating the same resource to maintain the execution of each element would lead to a decrease of customization efficiency to some extend. In order to solve this problem, this paper present an evaluation method to value the importance of each element in SN. After accumulating customer requirements and customization plans to a certain amount, we will apply the dynamic evaluation process to estimate the importance of each element in SN. Combining the results of dynamic evalution with the static structural characteristics of elements, we develop strategies to revise the SN for better performance of customization.3. The SN should not be a static network. Customer requirements show differences with the passage of time, so that the usage of SN elements will be changed along with the variation of requirements. As a result, both bi-lateral patterns that extracted from requirements ant their customization plans a s well as the elements in SN should be evoluted over time. Both customers and the maintenance engineer of SN may get to know the requirement and the usage of SN better by understanding the rules of evolution. When considering the cocept of time in the process of customization, bi-lateral patterns as well as the used elements of SN may set up connections among different temporal intervals in different types. In this paper, we design different types of evolution to distinguish the changes of bi-lateral patterns and SN elements between two consecutive time periods, and then use these mappings of connections to set up evolution networks and analyse them.We conclude from experiments and stasistics that:(1)compare to traditional approaches, our BPSC algorithm may lead to higer efficiency of customization;(1)after applying the strategies that generated from evaluation results of SN elements, we get better performance of customization by using revised SNs, especially for the time that has been compressed a lot;(3)we last generate several evolution networks for requirement patterns, service patterns as well as elements of SN with three different time span requirement sets and then the distributions of evolution types of each evolution network are further analyzed.