Optimization of the multiple retailer supply chain management problem

Authors:
Caio Soares;Gerry Dozier;Emmett Lodree;Jared Phillips;Katie Nobles;Yong won Park
Affiliations:
Auburn University, AL;North Carolina A&T State University, Greensboro, NC;Auburn University, AL;Auburn University, AL;Auburn University, AL;Auburn University, AL
Venue:
Proceedings of the 46th Annual Southeast Regional Conference on XX
Year:
2008

Citing 4
Cited 1

Genetic Algorithms in Search, Optimization and Machine Learning

Genetic Algorithms in Search, Optimization and Machine Learning
On the State of Evolutionary Computation

Proceedings of the 5th International Conference on Genetic Algorithms
Evolutionary Computation

Evolutionary Computation
The particle swarm - explosion, stability, and convergence in amultidimensional complex space

IEEE Transactions on Evolutionary Computation

Stability of the supply chain using system dynamics simulation and the accumulated deviations from equilibrium

Modelling and Simulation in Engineering

Quantified Score

Hi-index	0.00

Visualization

Abstract

With stock surpluses and shortages representing one of the greatest elements of risk to wholesalers, a solution to the multiretailer supply chain management problem would result in tremendous economic benefits. In this problem, a single wholesaler with multiple retailer customers must find an optimal balance of quantities ordered from suppliers and acceptable lead time costs, while taking into account limiting factors such as the time each retailer will wait for a backorder. The following four evolutionary computations (EC) are utilized to find a solution: evolutionary programming (EP), genetic algorithms (GA), particle swarm optimizers (PSO), and estimation of distribution algorithms (EDA). In addition, problem-specific modifications to each are created. Of the 32 attempted algorithms, the following proved to be best with respect to the client-mandated test-suite: Probabilistic Dual-Topology Full-Model PSO, Star-Topology Full-Model PSO using dynamically-adjusting learning rates, Out-of-the-Box Star-Topology Full-Model PSO, and a Gaussian-based Star-Topology Full-Model PSO with the Constriction Coefficient. A secondary test-suite was also developed to test the effectiveness of the best algorithms on the problem. With respect to the client-mandated and the developed test suite's fitness threshold and maximum number of function evaluations, the best algorithm had an 87% and 90% success rate, respectively. Considering the flexibility and high performance of the solution and the generality of the problem, these results represent a significant contribution to commercial wholesaling.