Ant colony algorithm, which is mainly used to solve discrete combination optimization problems, is a swarm intelligence algorithm. There are a lot of continuous optimization problems in the engineering, discrete ant colony algorithm for discrete combination optimization is extended to continuous ant colony algorithm, which has caused widespread concern. Now, there have been several continuous domain models of ant colony algorithm. This paper mainly research and improve on the ant colony optimization for continuous domains model, and the improved algorithm for parameters optimization of support vector machine.The ant colony optimization for continuous domains(ACOR) needs long computational time and easily traps into local optimal solutions, so artificial bee colony based ant colony optimization for continuous domain algorithm(ABC-ACOR) was put forward to solve the problems. Firstly, the introduction of an alternative mechanism instead of the original sortbased selection method to guide the solution chosen, saved the computational time and kept the diversity as long as possible. Secondly, combining artificial bee colony search strategy to improve the algorithm’s global search ability, further reduced the computation time and improved the accuracy of solutions. The proposed ABC-ACOR had been evaluated on a large of test functions. The experimental results show that ABC-ACOR performs better than some existing continuous ant colony optimization algorithms.The hybrid ant colony optimization for continuous domains(HACO) easily traps into local optimum solutions and converges slowly, so pheromone based adaptive hybrid ant colony optimization for continuous domain algorithm(QAHACO) was put forward to solve these problems. Firstly, a new approach was proposed to update the solutions, which made solutions pheromone itself evaporate, broaden search range and improved the global search ability. The adaptive pheromone evaporation rate was introduced in order to reach a better balance between convergence speed and convergence accuracy. Secondly, an information sharing mechanism was proposed, combining the average distance between the chosen solution and all other solutions and the distance between the chosen solution and the optimal solution found, further improved the convergence speed. Through simulation on test function, the results show that, compared with ant colony optimization for continuous domains and its improved algorithm, the accuracy of QAHACO algorithm improved significantly, convergence speed of QAHACO algorithm has certain advantages.Support vector machine is a small sample machine learning method. Owing to simple, low computational complexity and good robustness, the algorithm has become current research focus. In order to solve the difficult problem that support vector machine parameter optimization methods fell into local optimal solution in different degree,ABC-ACOR and QAHACO algorithm were applied to support vector machine parameter optimization. Penalty factor C and kernel function parameter σ of support vector machine were taken as the variables of the optimization objects function, and the classification accuracy of support vector machine was used as the fitness function value. Tests on five standard UCI datasets show that, compared with particle swarm optimization based support vector machine parameter optimization and genetic algorithm based support vector machine parameter optimization, ABC-ACOR based support vector machine parameter optimization and QAHACO based support vector machine parameter optimization obtained higher classification accuracy. Therefore, continuous ant colony optimization based support vector machine parameter optimization method is feasible. |