Research On Service Composition Approaches Based On Uncertain QoS-aware

With the rapid development of the Internet,more and more resources are released and utilized in the form of Web services.Meanwhile,the explosive growth of services has led to many similar or identical services on the Internet.How to select high-performance services to build value-added applications has become a research hotspot in academia and business community at home and abroad.Generally,Quality of service(QoS)has been considered as a significant criterion for measuring the performance among functionally similar or identical Web Services.Existing QoS-aware service composition methods include integer linear programming,mixed linear programming,and artificial intelligence algorithms,etc.However,these methods only consider certainty QoS,while ignoring the QoS uncertainty caused by the instability in the Internet environment,which resulting in inaccurate service selection results.How to solve the problem caused by uncertain QoS has become a very challenging research topic.On the one hand,existing QoS-aware service composition researches only consider the QoS performance of Web services at the composing time,while ignoring the success of service composition largely depends on the ability of different Web service components to maintain long-term stable quality of service.The composite service with the best current QoS performance is not necessarily the best after a period of time.On the other hand,the Internet is always in a high dynamic environment.When visiting services at different times,the QoS values may be different.The continuous change of QoS will break the optimal state of the original system components,and it is need to dynamically adjust to track the optimal state of system.In order to solve this problem,we need to find the best service composition solution for users in the case of continuous changes in QoS.This article studies the problem of service composition optimization based on uncertain QoS from the aspects of uncertain QoS modeling,service composition modeling and algorithm analysis.The main contributions of this article are as follows:(1)A service composition method based on uncertain-long time series(ULTS)was proposed.Firstly,the method is based on the user's access rules to the service,and the long-term change of quality of service was constructed as an uncertain-long time series model.The model can accurately describe the real QoS access record of the user to the service over a period of time.Secondly,an improved genetic algorithm T-GA was proposed.The tournament selection strategy is used to replace the roulette selection strategy in the basic genetic algorithm.Randomly select some individuals from population and insert into offspring.Randomly select some individuals from population with no duplicate and operate one-point crossover.The individuals generated by select and crossover operation constitute a new population.Then,select some individuals from the new population to perform the mutation operation.The proposed algorithm not only maintains population diversity,but also accelerates the convergence speed.(2)A dynamic service composition method based on directed search strategy(DSS)was proposed.The method considers that the QoS values of the service are different at different times in the dynamic network environment,and a new service composition method based on directed search strategy was proposed.The method contains two mechanisms,one used when environment change is detected,and the other used in each generation.The first mechanism is used to reinitialize the population after an environmental change.After a change occurs,part of the population is reinitialized with individuals generated in the predicted regions where the new non-dominated solutions may be located.In addition,the rest of the individuals are generated by performing a local search along the orthogonal directions of the predicted Pareto set(PS),aiming to enhance the diversity of the population.Secondly,the second mechanism aims to improve the speed of convergence to PS.At every generation,some individuals generated around the PS region predicted using the history information of the PSs are inserted into the population to speed up the convergence.The combination of the two mechanisms not only keeps the diversity of the population but also accelerates the convergence rate.(3)A large number of experiments are carried out on the data set WS-Dream.The experimental results show that the service composition method based on uncertain-long time series can effectively solve the problem of uncertain QoS-aware service composition.The proposed T-GA is superior to the E-GA algorithm in optimization results and stability,and the execution speed is improved by almost one time.The dynamic service composition method based on directed search strategy can respond to the continuous changes of QoS in time.Moreover,the non-dominated optimal solutions have a good diversity.Finally,the service composition approaches based on uncertain QoS-aware is analyzed and summarized,and the next research direction is given.