Research On Service Pattern Mining And Evolution For Exploratory Service Composition

With the arrival of the service trend,technologies such as Web services and service computing have been formed and developed one after another.While many services bring convenience to people's lives,there are also problems that users cannot identify business needs in advance,the services themselves are diverse,and the service composition efficiency is low.Exploratory service composition supports flexible service orchestration under controllable guarantees and can solve the problem of uncertain user needs.In the process of users'gradual exploration,they have accumulated successful experiences and failed lessons.If these experiences and lessons can be saved as reusable service patterns,the efficiency of service composition can be effectively improved.The service patterns are not static.With the passage of time,service resources and user needs change continuously,so the service patterns need to evolve accordingly to ensure their effectiveness.Therefore,how to construct the service pattern model,and how to mine and evolve them efficiently is the focus of this paper.The main work and contributions of this paper include:First,a service pattern mining method is proposed to mine the successful service composition processes and the failed service composition processes respectively.Firstly,a service pattern model is established to express the service pattern formally.Afterward,the gSpan algorithm and the failure service pattern mining algorithm(FSPMA)proposed in this paper are used to efficiently mine the two service composition processes respectively.The success service patterns and the failure service patterns are obtained.Among them,the FSPMA algorithm extends the gSpan algorithm,focusing the mining on the failure track of the failed service composition process,and improves the mining efficiency of failure service patterns.Experiments show that,compared with the gSpan and TKG algorithms for mining the entire failed service composition process,the mining efficiency is increased by 43%and 50%respectively.Second,a service pattern evolution method is proposed,which evolves the success service patterns and the failure service patterns respectively.This method applies a timeliness-based incremental service pattern evolution algorithm proposed in this paper to incrementally mine the new service composition processes.The algorithm combines the time decay formula to obtain the latest service usage with less time and resource costs,which efficiently achieves the evolution of the service patterns.Among them,the incremental method can effectively improve the efficiency of service patterns evolution.The experiments show that,compared with the FSPMA and gSpan algorithms,the running time of the algorithm is reduced by 40%and 67%respectively after using the incremental method.Third,the obtained service patterns are applied to service recommendation to verify the effectiveness of the service pattern mining and evolution method proposed in this paper.Experiments show that,compared with the traditional service recommendation method,the service recommendation method using the service patterns has effectively improved the precision rate,recall rate,and F1 value.Fourth,the service pattern mining and management system is designed and implemented,which mainly includes three modules:service pattern mining,service pattern management,and service recommendation.Among them,the service pattern mining module realizes the rapid mining of two service composition processes and generates the corresponding service patterns;the service pattern management module manages and evolves the service patterns;the service recommendation module uses the service patterns generated by mining and evolution to recommend services and assist users to solve exploratory problems.