DBSCAN-Based Adaptive Bacterial Foraging Algorithm Optimization

Bacterial foraging algorithm as a burgeoning swarm intelligence optimization algorithm, due to its stability and good global search ability, has been widely used. However, the traditional bacterial foraging algorithm convergence speed is slow, when dealing with high dimensional function and single step caused the contradiction between the convergence accuracy and convergence speed. Subsequently the adaptive bacterial foraging algorithm has been put forward. In the algorithm bacteria initially using large step to convergence fast and along with the objective function value reduced, bacteria step length based on the linear adjustment function is diminishing at the same time. The adaptive bacterial foraging algorithm, to some extent, solves the problem of chemotactic step size choice in bacterial foraging algorithm and subsequently accelerate the convergence rate. But along with the decrease of bacterial cost function value, the original chemotactic step size adjust function is liable to make chemotactic step size minimum, leading to the algorithm premature. In this article, adaptive bacterial foraging algorithm based on density-based spatial clustering of applications with noise (DBSCAN-ABFA) is designed, with the purpose of avoiding the algorithm premature by changing the chemotactic step size adjust function of the labeled core points bacterial according to DBSCAN, and the improved chemotactic step size adjust function can reduce the shrink rate of step size, so the algorithm premature is ultimately avoided. Specific work is as follows.(1)Bacterial foraging algorithm is different from other swarm intelligence optimization algorithm, bacterial populations are always stable moving towards higher nutrition concentration area, species foraging process is also observed convergence process. This inspires us introducing the idea of density clustering to mark the eutrophication area bacteria in the process of bacterial populations foraging. Then we can reduce the shrink rate of the bacteria chemotactic step by changing the adaptive step adjustment function to ensure the bacterias’ mobility and avoid premature.(2)There will be a greatly improve time cost of the algorithm, if we clustering the bacterial population in each iteration of the bacterial foraging algorithm by DBSCAN algorithm. So some improvements were made in this paper, in the bacterial foraging progress most bacteria will converge in one class near the optimal solution, and the bacteria cost function value is close to each other in the class. So we can find the closest bacteria based on the current cost function value rather than traverse the entire population. The improved algorithm reduced the times of calculate the space distance, finally reduce the time complexity of mark the bacterial population.(3)The parameter X in linear adaptive step function is discussed, aiming at the problem of artificial selection X hardly and lack of reference. So an adaptive selection parameters lambda method was proposed. This new method choose a relatively reasonable λ for every bacteria according to it’s initial cost function value and no longer one X for the population. In addition some contrast experiments are designed to verify the feasibility and effectiveness of the algorithm.