Research On Trusted Service Composition Method Based On Collaborative Awareness

With the development of economic globalization and the advancement of information technology,the service economy has become an important fundamental element of the global economy.In the service economy,service portfolio,as an important means of service innovation and value-added,is increasingly receiving attention from academia and the industry.Service composition refers to the combination of two or more services in a specific service supply chain or service ecosystem in a specific way to form new service products or solutions.Credibility is an important basis for users to choose and combine services.When facing numerous service providers and service nodes,users need to evaluate and compare the credibility of each service.Therefore,service trustworthiness is one of the prerequisites for service composition,and the reliability of service data to a certain extent ensures the reliability of the composition scheme results.There are significant differences in the selection of reference indicators in current research on service credibility evaluation,and there is no widely accepted standardized method to evaluate the credibility of services.Most studies are based on static data for evaluation,which cannot reflect the real-time status and social relationships of services.Most service composition systems do not consider fault tolerance and error tolerance.By interacting and sharing data information,sensing the collaborative characteristics between services,service nodes can reduce uncertainty issues during the execution process,ensuring the stability and reliability of the entire composition.Combinatorial optimization algorithms usually use genetic algorithms to solve the problem of large exhaustive search space,but these algorithms have their limitations,such as easy to fall into local optimal solutions,slow convergence speed,poor performance in large-scale scenes and other problems.This thesis focuses on the research of trusted service composition based on collaborative perception,aiming to ensure the credibility of services and the efficiency of composition schemes in large-scale service scenarios.The main tasks are as follows:(1)In response to the fact that traditional service trust evaluation usually only relies on direct data quality indicators,which cannot accurately capture the complex and constantly evolving nature of modern services,this thesis proposes a service credibility evaluation method PSO-STC.This method divides the trust features into direct trust and indirect trust,and uses the random forest algorithm to determine the weight of four trust features: friend,organization,reward and punishment feedback,and cooperation.The mutual information mechanism reduces model redundancy,and the particle swarm optimization algorithm ensures the accuracy of the results.Finally,aggregate direct and indirect trust to obtain the overall credibility of the service.(2)Traditional service composition methods face the following challenges when solving composition problems: QoS attributes are suitable for idealized environments,but in practical applications,the process is complex and involves a large number of types of services,which is prone to frequent unexpected occurrences and difficult selection.In order to solve these problems,this thesis proposes a combinatorial optimization method S-DDL-SC based on deep reinforcement learning.Firstly,in order to ensure the feasibility and stability of service composition,collaborative attributes such as technology,quality,social similarity,and capability matching between services are taken into account.Based on QoS attributes,a collaborative model is constructed to ensure the degree of collaboration between services and reduce uncertainty.Secondly,for problems in large-scale service composition scenarios,optimization learning methods using deep recurrent neural networks can better capture the interaction relationships between different services,making service composition more intelligent,accurate,and efficient.(3)This thesis fully confirms the effectiveness of the proposed method by using datasets such as Sigcomm-2009 and QWS.In order to deeply analyze the trust indicators of service evaluation,this thesis uses clustering method to successfully determine the trusted Decision boundary,and further analyzes the highest accuracy of trust evaluation by setting different trust thresholds.At the same time,in order to verify the effectiveness,adaptability and scalability of the combinatorial optimization algorithm in large-scale scenarios,this thesis also designed three groups of contrast experiments,and compared with PD-DQN(A dueling Deep Q-Network with prioritized replay)based on deep reinforcement learning algorithm and QCN(Q-learning for Adaptive Service Composition)based on strong chemical learning algorithm,to verify the efficiency of S-DDL-SC algorithm in service combinatorial optimization algorithm.Compared to traditional methods,the method proposed in this thesis can maintain stable and efficient QoS attributes and collaboration levels while ensuring the trustworthiness of services when solving service composition problems,ensuring the successful execution of tasks.