Measuring Efficiency Intervals in Multiple Groups with Privacy Concerns

Abstract

The organization is willing to take data from the other organizations for the better analysis, when the self-accumulated data only gives the limited findings for instance in case of focusing on a specific small group. The multi-accumulated data can be available for the analysis, if the different several organizations accumulated similar data independently. However, the organizations need to care for privacy since the original data could include a kind of personal information. Under such a privacy concern, this paper proposes the method for efficiency evaluation with multi-accumulated data based on Data Envelopment Analysis (DEA), which is a non-parametric technique to measure efficiency of a decision making unit (DMU) relatively in a group. The efficiency interval of a DMU in a group by referencing DMUs in another group as well as DMUs in its own group is obtained, even if the group cannot access to the original data of another group. Instead, the group takes the information of the efficient frontier of another group denoted as the weight set, from which the group cannot guess the original data of another group. As a result, three kinds of efficiency intervals for a DMU are obtained: the efficiency in its own group, that in another group, and that in the integrated group. Comparing them can give us a rich and useful information on the DMU from wide viewpoint.