A comparison between genetic algorithms and the RAND method for solving the joint replenishment problem

The purpose of this paper is to compare the performance of genetic algorithms (GAs) and the best available heuristic, known as the RAND, for solving the joint replenishment problem (JRP). An important feature of the JRP which makes it suitable for GAs is that it can be formulated as a problem having one continuous decision variable and a number of integer decision variables equal to the number of products being produced or ordered. Experiments on randomly generated problems indicate that GAs can provide better solutions to the JRP than the RAND for some problems, and at worst can almost match the performance of the RAND from a practical point of view for the rest of the problems. GAs never converged to solution with a total cost of more than 0.08% of the total cost of the RAND for 1600 randomly generated problems. In addition, GAs have the advantages of: (i) being easy to implement (e.g. less than 200 lines of code); (ii) having a code which is easy to understand and modify; and (iii) dealing easily with constrained JRPs which are neglected by most of the available methods including the RAND, in spite of their importance in practice.