Improved Modified Bacterial Foraging Optimization Algorithm to Solve Constrained Numerical Optimization Problems

Abstract

This paper presents an improved version of the modified bacterial foraging optimization algorithm to solve constrained numerical optimization problems. Four mechanisms are added: (1) two swim operators, one to favor the exploration and another one to focus on the exploitation of the search space, where a dynamic mechanism is considered to deal with the stepsize value, (2) a skew mechanism for a more suitable initial swarm where bacteria are divided in three groups, two of them close to the boundaries of the search space and one distributed in all the search space, (3) a local search operator and (4) a decrease in the usage of the reproduction step to deal with premature convergence. 60 well-known test problems from two benchmarks are solved along three experiments. The first experiment aims to provide preliminary evidence on the suitable behavior of the new mechanism added. The second experiment provides an in-depth comparison of the new version against its previous one based on final results and four performance measures. The third experiment compares the performance of the proposed algorithm against five state-of-the-art nature-inspired algorithms designed to deal with constrained continuous search spaces. The results show that the proposed algorithm clearly provides a better performance against its predecessor by increasing its ability to reach the feasible region and generating better solutions, while obtaining a competitive performance against those compared state-of-the-art algorithms.