Allocation efficiency is an important performance index to measure the quality of the allocation algorithm. In order to compute the efficiency, the volume of the subset of attainable moments must be solved. The efficiency of the redistributed pseudo inverse (RPI) algorithm depends on the choice of the pseudo-inverse matrix. The subset of attainable moments of RPI is a complex non-convex polyhedron. By analyzing two-dimensional and three-dimensional allocation problems with a “micro-element” method, here we propose an approximate calculation algorithm to compute the volume of the non-convex polyhedron. In order to improve the allocation efficiency of RPI, genetic algorithm is used to find the best pseudo-inverse matrix. The simulation results show that the best pseudo-inverse matrix can be easily chosen by the proposed method and the high allocation efficiency is achieved.