DEA-based indicator for performance improvement

This paper concerns an active indication method on data envelopment analysis and its experimental application. While DEA works to evaluate efficiency for the decision making unit with multiple input and output, there are still problems as follows: It is not easy (1) to view each DMU's position from bird-eye, (2) to catch the relationship between a DMU and its dominant set, and (3) to get an improvement direction from the current DMU position to the dominant set. To solve these problems, this paper clarifies the visualization procedure for indicating performance improvement and applies it to Japanese prefecture evaluation and verified in a two-input/single-output problem and a single-input/three-output problem as follows: (1) data normalization works well under the assumption that returns to scale is constant, (2) logarithmic transformation is useful in the case that the data range is strongly biased, and (3) viewpoint in three dimensional graph should be changed for confirming the direction.