To design a new variant of Shuffled Frog Leaping Algorithm in which memeplexes formation is
modified with new strategy.
Shuffled frog leaping (SFL) is a memetic meta-heuristic algorithm that inherits the features of two other
algorithms. Its intensification component of search is similar to Particle Swarm Optimization while the inspiration for
diversification is inherited from the global exchange of information in Shuffled Complex Evolution. Basic variant has
been applied to solve many optimisation problems. SFLA suffers with slow acceleration rate.
To propose a robust hybrid SFLA that accelerates convergence.
Two modifications are proposed in the structure of basic SFLA. Firstly, memeplexes formation is modified to
handle continuous optimization problems. Secondly, in basic SFL algorithm the position of worst frog is improved by
moving it towards the best frog in the respective memeplex, with the progress of execution, the difference between best
and worst frog position reduces; there may be more chances to trap in local minima. With an aim to improve convergence
and avoiding trapping in local optima a parent centric operator is embedded in each memeplex while performing a local
search. The proposed algorithm is named as PC-SFLA (Parent Centric - Shuffled frog leaping algorithm)
The improved efficiency of PC-SFLA is validated on a robust and diverse set of standard test functions defined in
CEC 2006 and 2010 and further its efficacy is verified to optimize the total cost of Supply chain management of a system.
Non-parametric statistical result analysis demonstrates the efficiency of the proposal.
PC-SFLA performed better than PSO, DE, PESO+, Modified DE, ABC and SFLA at 5% and 10% level of
significance where as at par with Shuffled-ABC for g01-g07 functions of CEC 2006 in terms of NFE’s. Similarly, PCSFLA performed better than SaDE, SFLA, CMODE at both level of significance (5% & 10%) and at par with MPDE in
terms of mean function value for 17 problems taken from CEC 2006. Further PC-SFLA is investigated on a set of 18
problems from CEC 2010 and Wilcoxon signed ranks test is performed at 5% level of significance. PC-SFLA performed
better than SFLA and CHDE and at par with PESO. The computational results present the competency of the proposed
method to solve quadratic, nonlinear, polynomial, linear as well as cubic functions efficiently. The simulated results
shows that the proposed algorithm is capable of solving mix integer constrained continuous optimization problem
efficiently.