| Popis: |
An essential part of a Service-Oriented Architecture (SOA) is interface of the service, which functions as an agreement between the one who provides the service and the clients. In order to accommodate changing requirements, these interfaces are often updated. However, a web service's subscribers' systems are frequently impacted by modifications to the interface. As a result, it's critical for users to assess danger of utilizing particular service and equivalence its development to rest of services that offer the similar functionality to minimize effort required to modify their applications in subsequent releases. Furthermore, foreseeing interface changes may assist online service providers in more effectively managing their resources (such as programmers' accessibility, strict timelines etc.) and scheduling necessary maintenance tasks to raise standard of their services. In this article, we suggest using artificial neural networks-based machine learning to forecast how the architecture of Web services interfaces will change over time. In order to achieve this, we gathered training data from six Web services' quality indicators of earlier releases. The validation of our prediction approaches reveals that, with a very less deviation rate, predicted metrics values for the various releases of six SOA based services, such as number of operations, were extremely similar to the ones that were expected. Additionally, with an average precision and recall more than 85%, the majority of quality concerns of examined Web service interfaces were correctly anticipated for subsequent releases. The study done through working developers demonstrates the value of prediction methods for both service consumers and providers. |
