HYBRID BACTERIAL FORAGING WITH FIREFLY ALGORITHM

Abstract

In artificial intelligence, an evolutionary algorithm (EA) is a subset of evolutionary computation, a generic population-based metaheuristic optimization algorithm. An EA uses mechanisms inspired by biological evolution, such as reproduction, mutation, recombination, and selection. Candidate solutions to the optimization problem play the role of individuals in a population, and the fitness function determines the quality of the solutions. EAs are well-known optimization approaches to deal with nonlinear and complex problems. EAs often perform well approximating solutions to all types of problems because they ideally do not make any assumption about the underlying fitness landscape. The computational complexity depends on the Fitness function evaluation. One such example of EA is Bacterial Foraging Algorithm. Bacterial Foraging Optimization (BFO) [1] is a bio-inspired optimization technique, proposed by K.M. Passion. It is based on the Escherichia Coli (E. Coli) bacteria’s foraging behavior i.e. the food seeking and reproductive behavior of bacteria. The classical BFO has three main mechanisms, viz, chemo taxis step, reproduction step and elimination dispersal step. The chemo taxis gives local optima whereas the reproduction, mutation and elimination-dispersal gives the global optima. Another nature inspired metaheuristic algorithm is the Firefly Algorithm (FA). It was developed by Xin- She Yang in 2008 at Cambridge University [19]. In this, the search algorithm is inspired by the social behavior of Fireflies. There are two important issues in this algorithm, namely, variation of light intensity and formulation of attractiveness. This thesis presents the hybrid of BFO with FA. Two modifications are made to the original BFO algorithm. Firstly, the position of bacteria in Bacterial Foraging Algorithm (BFO) is updated after all fitness evaluation calculations rather than each fitness evaluation calculation in chemo taxis step. Secondly, the bacteria positions are updated according to the position updation equation of Firefly Algorithm (FA). This step is defined as Mutation in the proposed algorithm. In this way, more accurate values of global optima are obtained using BFO and fast convergence is ensured using Firefly Algorithm. The technique has been applied on various benchmark functions to validate claims and results are compared with traditional BFO and FA. The results show that the proposed technique gives efficient results compared to the traditional algorithms.