Data envelopment analysis (DEA) is an useful performance assessment tool, which is able to consider multiple inputs and outputs simultaneously while requiring neither a priori weights nor a pre-specified functional form. To conduct a successful DEA study, data quality plays a key role but has not drawn much attention. Outlier detection not only identifies the suspicious data point to avoid erroneous conclusion, but also possibly leads to the discovery of unexpected knowledge. The objective of the paper is to develop a comprehensive data filtering scheme to DEA. We first examine several preliminary outlier detection procedures in DEA and discuss some unsolved issues. Then we develop the inefficient outlier detection which is yet unsolved. Besides providing a solution, this method is funded on the ground of DEA theory.
|出版狀態||Published - 1 十二月 2006|
|事件||36th International Conference on Computers and Industrial Engineering, ICC and IE 2006 - Taipei, Taiwan|
持續時間: 20 六月 2006 → 23 六月 2006
|Conference||36th International Conference on Computers and Industrial Engineering, ICC and IE 2006|
|期間||20/06/06 → 23/06/06|