Research On Web Service Composition Selection Strategy Considering QoS Correlation

At present,Service Oriented Architecture(SOA)has become the main framework of software engineering system,and Web service has become the implementation standard of SOA architecture.As a single Web service has gradually been unable to meet the needs of users,it is necessary to combine services to achieve complex business functions.As the functions of services become more and more complex,the number of services required in service composition is also increasing.For the same function,different candidate sub-services can be selected to form different service composition schemes,but their performance is different.And services are not necessarily independent of each other.There may be competition or cooperation in one way or another.For example,the composite service formed by the cooperation of the two services will be cheaper and less response time than that provided separately.To solve these problems,this paper first uses Jaccard similarity and cosine similarity to mine the similarity relationship between services,then uses common neighbor and resource allocation to mine the composable relationship,calculates the interface correlation based on the composable relationship and potential composable relationship,calculates the statistical correlation based on historical data analysis,and publishes the service according to the QoS value of the composite service and the service release.The quality correlation is calculated by comparing the QoS values provided by the authors.Finally,integer programming is used to solve the optimization model of multi-objective composite services considering correlation.This paper designs a banking business system with four structures: sequence,cycle,branch and parallel,and adds <link> structure to the parallel structure,which makes the execution of services in the process dependent and the overall structure more complex.NSGA-? is a multi-objective optimization algorithm using Pareto method.It is a classical algorithm in multi-objective optimization.On the basis of genetic algorithm,a fast non-dominant sorting method is used to allocate fitness for each chromosome,and then hybridization,mutation and iteration are carried out.However,when considering the correlation,the effect of the algorithm is not ideal.Based on the NSGA-? algorithm,this paper proposes a method of reallocating fitness,considering the relationship between offspring and parents after adding correlation and the change of one-dimensional QoS of a service after hybridization,and updating their fitness.The results show that the proposed algorithm has better convergence and convergence speed,and can obtain better approximate optimal solution than the traditional NSGA-? algorithm.