To decompose an unbalanced multi-stage logistic system to multipleindependent single-stage logistic systems, a new notion of parameterized interface distribution ispresented. For encoding the logistic pattern on each stage, the Pruefer number is used. With theimproved decoding procedure, any Pruefer number produced stochastically can be decoded to a feasiblelogistic pattern, which can match with the capacities of the nodes of the logistic system. Withthese two innovations, a new modeling method based on parameterized interface distribution and thePriifer number coding is put forward. The corresponding genetic algorithm, named as PIP-GA, can findbetter solutions and require less computational time than st-GA. Although requiring a little moreconsumption of memory, PIP-GA is still an efficient and robust method in the modeling andoptimization of unbalanced multi-stage logistic systems.
