Research On Key Technologies Of Semantic Web Dynamic Service Composition

As the rapid development of the Internet, a new digital era has coming to us. And then, all kinds of online application and network services have been getting more and more popular. Moreover, these applications and services will be in politics, economy, military, education, science, technology, culture, business, and other fields. As the real world, people’s demands of the Internet applications and services are always constantly changing, therefore, the applications and services supplied by the Internet should change as well. After all, this is also a kind of people-oriented ideology.As a new kind of distributed computing model, web services have obtained lots of concerns from industrial community and academic circles in recent years. Generally, a single web service with limited function usually cannot meet the multidimensional needs of practical applications, so how to form a new, stable, reliable and multi-functional DWSC (dynamic web services composition) which are able to support the QoS (quality of service) and satisfy web users’various requirements has become the continuous highlight in this research field.(1) Semantic web dynamic service composition model (SDCM)According to the characteristics of dynamic web service structure and active calculation, a semantic web dynamic service composition model is proposed in this paper, by which the purposes of the service requests are discovered and identified so as to live up to the requester’s service quality and service level, and the services are inquired and checked out from the service registration database to obtain the required services through binding their URI with service acquiring techniques.(2) Selective ensemble algorithm based on service classificationSDCM is based upon web services classification. AODE (averaged one-dependence estimators) algorithm is a kind of typical improved algorithm based on a?ve Bayes, which gets much attention from the international machine learning field. CEM (cross-entropy method) is a global random search algorithm for combination optimization problems, which is put into use for many classic NP-complete problems. CESAODE (cross-entropy method for selective AODE) was proposed based on cross-entropy method to improve the performance of Web services classification. The effective performance of CESAODE was then shown using a set of classification problems with UCI datasets based on Weka platform, which indicated that CESAODE was superior to other classification algorithms like AODE.(3) A real-time classification method based on multitask web servicesThe traditional real-time classification algorithms mostly focus on only single instance, namely, in a certain period, only one sample can be processed, meanwhile, others are waiting in the buffer. The whole sample set was just locally optimized, because a part of samples in the sample set were not evaluated sufficiently when the process was interrupted. A more flexible and effective real-time classification method based on T-distribution P-value for dynamic multitask web services is proposed in this paper, by which the whole sample set would get globally optimized because the most invaluable samples in the sample set had got evaluated before the process was interrupted.(4) A dynamic niche-based self-organizing learning mechanism for dynamic web services composition optimizationIn the SOA (service-oriented architecture), web users’demands are various and dynamic, yet the current studying on intelligent evolutionary algorithms are always carried out for static problems. These static-oriented evolutionary algorithms are apt to loss the ability to adapt to the environment when they approach to the end, and can not track the extreme points in solution space. Thus the algorithms are unable to solve the dynamic problems efficaciously. Thus, a dynamic niche-based self-organizing learning algorithm based on semantic web services was proposed in this paper. In which, a dynamic learning mechanism based on 0-1 coding method was carried out for the first time, and the individuals involved in this algorithm are able to adapt the dynamic environments consisting of web users’demands through active learning. the algorithm proposed here show a strong robustness in the comparative experiments, whose dynamic search capabilities are far superior to other search methods.(5) A model for detecting and processing web service unavailabilityThe usability of web service is restrained by services unavailability. With the perspectives of dynamic detection and effective disposition for service unavailability, a web services unavailability detection mechanism was designed to realize the effective compromise for integrity, accuracy and efficiency of the detection and adapt to the dynamic environment. Meanwhile, a fault-tolerant processing mechanism based on particle swarm optimization and c-mean clustering method was designed to set the boundary value of the service process’s anomalous state and reshuffle and optimize the services composition consist of parts of unenforced services after the initial process SLA(service-level agreement) was met.(6) The forecast of QoS (quality of services) demand and evaluation model for SDCMThe web QoS consists of a group of service’s non-functional attributes, and it rest with each sub-unit’s QoS. SDCM satisfied the web users` service requirements in the process of running through dynamically discovering, selection and binding component services according to the run-time dynamic environments and users’service requirements. A kind of relationship matrix encoding method based on service selection characteristic is given in the paper, by which one encoding method can symbolize more than one path, and also can stand for dynamic repetitive schedule and web service recirculation path, and the ability for symbolizing various services composition is improved. Finally, a QoS evaluation algorithm for the process of SDCM is designed and its availability, effectiveness, adaptability are tested becomingly by a number of experiments.