Abstract: Decision makers are often faced with how to allocate a certain amount of resources among peer Decision Making Units (DMUs) effectively in practice. Considering DMUs as systems, this paper focuses on the resource allocation with considering the system structure consisted of a series of parallel subsystems. Treating the allocated resource as a new independent input to DMUs, the allocations ensuring all overall systems and parallel subsystems pareto efficient simultaneously (efficiency values equal to 1) is obtained by employing data envelopment analysis (DEA), and then the efficient interval of allocation for the parallel subsystem is confirmed. The utility function is defined to reflect the satisfaction degree of parallel subsystems to the resource allocation. The final resource allocation model is proposed based on the principles of max-min fairness for satisfaction degrees and Pareto efficient. Finally, the validity of the proposed resource allocation model is demonstrated by the data of the three main industries of 31 provinces or regions in China in 2008. It has theoretical and practical guiding significance in some extent.

Key words: data envelopment analysis (DEA), resource allocation, parallel system, internal structure, max-min fairness