| Summary: | 碩士 === 逢甲大學 === 工業工程與系統管理學研究所 === 100 === Recently, Data Envelopment Analysis (DEA) has been widely used. However, the conventional DEA there exist a weakness that doesn''t allow stochastic variations in input and output, such as measure errors, enter errors and process variations. Also, the conventional DEA only can divide decision making units (DUMs) into two categories: efficiency and inefficiency. When multiple DMUs are efficiency, one can''t differentiate those efficient DUMs. In this study, we investigate the decision problem that the input and output may be stochastic, and use the chance constrained programming technique to convert the probability type of constraint into a deterministic equivalent. Also, in order to highlight the differences between the efficient DUMs, we combine the super-efficiency DEA with stochastic DEA into a stochastic super-efficiency DEA. Then, the proposed method is applied to the wafer fabrication multi-response quality problem. The measurement data which provided by a wafer fabrication company is used to analyze and assess the quality performance on the combination of the process parameters in the wafer fabrication. Finally, we use the change of alpha value to discuss the sensitivity analysis of the stochastic super-efficiency DEA. Results indicate that the developed approach can provide good solutions for stochastic multi-response quality problems under uncertain environments.
|
|---|
