Hybrid Intelligent Algorithm And Its Application In Optimization Problems

The optimization problems include the continuous optimization problems and the discrete optimization problems.As for multimodal function problem in continuous optimization problems and Non-Deterministic Polynomial(NP)problems in the discrete optimization problems,it is easy to fall into the local optimal solution when the general intelligent algorithms,if not improved and adjusted,are used to solve these problems.Therefore,this paper selects four most widely used intelligent algorithms on the basis of modification and improvement of the original algorithms to solve some complex continuous optimization problems such as continuous function optimization,discrete combinatorial optimization problems,such as Traveling Salesman Problem(TSP),clustering problems applied to hierarchical optimization,and restrained task scheduling,from which more ideal optimization effects can be drawn than the original algorithm and some other famous classical algorithm.The main contents of this paper include four parts as follows:(1)This paper proposes a glowworm swarm optimization algorithm based on the quantum behavior of elite learning for defects like low accuracy and prematurity in the process of function optimization when using glowworm swarm optimization algorithm and to solve continuous function optimization problems.In the stage of initial population,the algorithm uses initial population of Logistic mapping chaotic mechanism to improve the randomness and diversity of the initial population.In the adaptive dynamic step size search for the position of glowworm swarm,it will take quantum behavior strategy of random mechanism to update those who are not selected because of roulette law and reinitialize cross-border individuals to ensure each individual has different changes,which improves the probability of optimization.At last,it takes dynamic approximation learning strategy for the elite individual selected by evaluating fitness value,which greatly improves the local search ability of the algorithm avoiding falling into local optimum.Experiments on testing CEC2014 Benchmark functions show that the improved artificial glowworm swarm algorithm has such more advantages as higher rate of convergence and higher accuracy of solution than other optimization algorithms.(2)This paper proposes the Hybrid Immune Algorithm(HIA)for the problems of too slow convergence and closed competition appearing in the application process of basic immune algorithm when solving discrete optimization problems.The immune algorithm adopts the hybrid method,which is used for global search in the individual and local research executed by initial population of the greedy algorithm and de-crossing operator together in the chromosome.When updating individuals,it will improve the mutation operator in the basic immune algorithm and employ high-frequency mutation operator that works under the dynamic adaptive mutation probability.These strategies have improved the diversity of populations and are easier to find potential and more effective search directions,reducing blind search,which makes the offspring population explore in a more favorable direction and rapidly search solution region with higher quality and overcome the contradiction between depth mining and range scanning until the balance between them is reached.Results of experiments show that it can be seen that compared with other algorithms,HIA can find the more ideal global optimal solution and have more stable performance.which is an efficient algorithm for solving discrete optimization problems.(3)For the problems of low efficiency of large-scale TSP problems,the general solution is hierarchical method.Firstly,the clustering algorithm is used to transform the large-scale TSP problem into several small scale city clusters,and then the problem is regarded as a generalized traveling salesman problem(GTSP).This paper proposes a cloud model intrusion weed optimization algorithm based on Invasive Weed Optimization algorithm(IWO).This algorithm uses the cloud model intrusion weed optimization algorithm to lead the search of k-means algorithm,which makes the population have a clear direction in the evolutionary process improving the searching ability of the algorithm.In order to verify whether the proposed algorithm is correct and efficient or not,this paper makes experiments on three clustering problem in its experimental part,which proves the proposed algorithm has not only high accuracy and faster convergence speed,but also strong stability.(4)The essence of task scheduling is to realize the mapping from dependent sub sets to a set of processors and seek the best allocation scheme in order to obtain the minimized makespan.This problem is a restrained combinatorial optimization problem,which is generally an uncertain polynomial-time-hard problem(NP problem).In order to solve this problem better,this paper adopts Artificial Chemical Reaction Optimization Algorithm(ACROA)to simulate five main molecular operations in the process of chemical reaction,making the reactants to interact with each other to achieve the minimum enthalpy(potential energy)state,from which the Artificial Chemistry Reaction Optimization Algorithm for solving Job Scheduling(ACROAJS)problem is proposed.The experimental results show that compared with the other two solutions in the literature,the proposed algorithm in this paper improves the makespan in the grid computing environment,reducing the makespan by about 5.06%.In the conclusion of this paper,the four improved representative intelligent algorithm are summarized,and the further research directions are put forward.