Improved Fireworks Algorithm And Its Application In Web Service Composition

Swarm intelligence algorithm is an optimization algorithm inspired by biological activities or natural phenomena.This kind of algorithm searches the neighborhood space through the naive behaviors of individuals,and then updates the individual information combined with the information of the population to implement the evolution of the population.As a new swarm intelligence algorithm,fireworks algorithm has attracted much attention in the field of scientific research.The algorithm is mainly composed of three parts:explosion operator,mutation operator and the selection strategy.Explosion operator generates explosion spark to search fireworks neighborhood space,and adjusts explosion radius and explosion spark number according to individual fitness of fireworks,so that the global search ability and local search ability are balanced.Mutation operator can generate spark of mutation and increase the diversity of population.The selection strategy selects several individuals from fireworks and sparks to form the next generation fireworks population.Fireworks algorithm has been proposed for a short time.However,compared with the existing swarm intelligence algorithms,it has the characteristics of high explosiveness,population diversity and local coverage,and has been widely used in the fields of filter design,image recognition and fertilization.As a new algorithm,fireworks algorithm still has some problems such as slow convergence speed and unsatisfactory searching precision while confronting some practical problems.Therefore,it is of great significance to study the internal principle of fireworks algorithm,improve it and apply it to practical problems.This thesis studies the principle of basic fireworks algorithm,analyzes the shortcomings of existing algorithms,and puts forward several improvement schemes.First,the performance of the improved algorithm is tested on the benchmark function,and then the algorithm is applied to the solution of Web service composition optimization problem.This thesis is mainly divided into two parts:1.This thesis proposes an improved fireworks algorithm with Cauchy mutation operator(CFWA)to solve the problem that the basic fireworks algorithm and its improved algorithm are prone to fall into local optimization in the process of optimization.This algorithm introduces the Cauchy mutation operator and elite candidate set strategy on the basis of the original improved algorithm,which reduces the probability of the algorithm falling into the local optimum and improves its searching precision.First,the Cauchy mutation operator is used to mutate the optimal fireworks,which enhances the local optimization ability and the global optimization ability of the optimal fireworks,and increases the diversity of the population.Then the elite candidate set strategy is used to screen the explosive spark and the variable spark.This strategy is divided into directional spark mechanism and"survival of the fittest"mechanism:the directional spark mechanism avoids the over-concentration of the next generation population and increases the diversity of the next generation fireworks;"survival of the fittest"mechanism makes fireworks variation has a certain direction,leading fireworks to a better direction of variation.Experimental results show that this method has good performance.2.In the thesis,the improved Cauchy fireworks algorithm is applied to the problem of Web services portfolio optimization,and the NP-hard problem of Web services portfolio optimization is solved by taking advantages of the strong global search ability and high optimization accuracy of Cauchy fireworks algorithm.the Cauchy fireworks algorithm is firstly modeled discretely,then the performance of the algorithm is tested by simulating Web service data sets of different scales.The experimental results show that compared with the other three algorithms,the Cauchy fireworks algorithm can find a better optimal solution in a reasonable time and has good stability when solving the Web service portfolio optimization problem,especially when the data size is large.