Hybridized Optimization Algorithms Of Swarm Intelligence And Their Application

Single swarm intelligent optimization algorithm has the disadvantages of low accuracy, weak generalization ability and easy to fall into local optima in solving complex problems. While hybridized algorithm of swarm intelligent optimization can realize the complementary advantages and information value-added by combinating the differentiation, the complementary of intelligent optimization algorithms, and enhance the overall performance of the algorithm in solving the complicated optimization problem. The thesis mainly studies the improvement of particle swarm optimization algorithm and artificial colony algorithm. Some novel hybridization swarm intelligence algorithms are proposed by embedding intelligent optimization algorithm including chaotic optimization algorithm and differential evolution algorithm to improve the convergence speed and escape from the local optimum effectively. Then these algorithms are applied to solve the practical optimization problems.The main contents of this dissertation are as follows:(1) Improved chaotic particle swarm optimization algorithm and its applicationThe performance of particle swarm optimization (PSO) offen suffers the problems of slow convergence speed during the later period and trapped in local optima. In this thesis, a novel improved chaotic particle swarm optimization (ICPSO) is proposed by introducing chaos optimization algorithm and the judgment and processing mechanism of the premature convergence. The results of algorithm analysis and function optimization experiments show that ICPSO can enhance the convergence speed avoiding the later oscillation.The ICPSO algorithm is applied to select the best parameters combination of support vector machine (SVM) and dynamic fuzzy neural network (DFNN), respectively. Then the parameter optimized SVM and DFNN model are applied for coal and gas outburst prediction. Comparision experiment results show that ICPSO can efficiently obtain the optimal parameters combination, with higher precision, faster convergence and less iteration, which can improve the performance and prediction precision of SVM and DFNN model.(2) Chaotic differential evolution particle swarm cooperative optimization algorithmTo improve the performance of differential evolution particle swarm optimization, a novel chaotic differential evolution particle swarm cooperative optimization algorithm is proposed by introuding the chaotic optimization, the opposition-based learning and the information exchanging mechanism. In this algorithm, an initialization strategy based on the opposition-based learning is applied to diversify the initial individuals in the search space. All individuals are randomly divided into two sub-swarms, one sub-swarm searches via improved chaotic differential evolution, and the other one searches via ICPSO. The results of algorithm analysis and function optimization experiments further demonstrate that the algorithm not only effectively avoids the premature convergence, but also has rapid convergence speed, high solution precision, good searching ability and robustness.(3) Self-adaptive Tent chaotic artificial bee colony algorithmIn order to improve the disadvantage of the ABC algorithm effectively, such as the slow global optimal search speed, decreased species diversity, even the local optimum. Comparing with Logistic map, chaotic sequences produced by Tent map behave with global ergodicity and uniform, which are insensitive to initial value. Therefore, a novel self-adaptive Tent chaotic artificial bee colony algorithm (STO-CABC) is proposed to enhance the global convergence and avoid premature convergence. In the proposed method, Tent chaotic opposition-based learning initialization method is presented to diversify the initial individuals and obtain good initial solutions. Furthermore, the self-adaptive Tent chaotic searching is implemented at the zones nearby individual optimum solution. Moreover, the tournament selection strategy in onlooker bee phase is employed. The results of algorithm analysis and experiments further demonstrate that STO-CABC is better than the basic ABC in term of the convergence rate and precision, and provides excellent performance in dealing with complex high-dimensional functions. Finally, STO-CABC algorithm is employed to optimize the parameters combination of support vector regression (SVR) model for annual electric load forecaste. The results of comparision experiment show that this model can obtain the optimal parameters combination efficiently, and has good modeling and higher prediction accuracy.(4) Hybridization algorithm of Tent chaotic artificial bee colony and particle swarm optimizationThe ABC algorithm has effective global search ability (exploration) but poor local search ability (exploitation), while the PSO has powerful local search ability but poor global search ability. Therefore, a hybridization algorithm of chaotic artificial bee colony and particle swarm optimization (HTCAP) is proposed, which is combined by Tent chaos search, the opposition-based learning and the recombination operator. In this algorithm, an initialization strategy based on the chaotic opposition-based learning is applied to diversify the initial individuals in the search space. All individuals are divided into two sub-swarms, one sub-swarm searches via STO-CABC, and the other one searches via Tent chaotic particle swarm optimization (TCPSO). The global best solutions obtained by the STO-CABC and TCPSO are used for recombination, and the solution obtained from this recombination is given to the populations of the STO-CABC and TCPSO as the neighbor food source for onlooker bees and the global best, respectively. Experiments on the complex benchmark functions with high dimension, simulation results further demonstrate that, the algorithm not only avoids the premature convergence effectively, but also gets rid of the local minimum. By comparison with other hybrid algorithms based on the PSO and ABC, the proposed model has better global and local searching abilities. Finally, the HTCAP algorithm is employed to solve multiple sequence alignment. Compared experiment results show that the algorithm can solve multiple sequence alignment efficiently, and obtain better alignment results and more robust.(5) Hybridization optimization algorithms of swarm intelligence for intrusion detectionKernel principal component analysis (KPCA) is a nonlinear feature extraction method, and the generalization performance and regression accuracy of support vector machine (S VM) depend on its parameters selection. A novel SVM model combining KPCA with swarm intelligence hybridization optimization algorithms is proposed for intrusion detection. In the proposed model, a multi-layer SVM classifier is adopted to estimate whether the action is an attack. KPCA extracts the features of initial data to reduce the dimension of feature vectors and shorten training time. In order to reduce the noise caused by feature differences and improve the performance of SVM, an improved kernel function is proposed by embedding the mean value and the mean square difference values of feature attributes in RBF kernel function. The four proposed hybridization algorithms of swarm intelligence are employed to optimize the parameters of SVM. By comparison with other detection model, the experimental results show that the proposed model has better detection performace and generalization.