Integrated comprehensive methodology based on DEA and PCA

Data Envelopment Analysis (DEA) is a kind of comprehensive evaluation method to rank and calculate the corresponding efficiency for each Decision Making Unit (DMU) by constructing a production frontier in multiple inputs and outputs system. For each DMU, we can use multiple variables to describe the characters of DMU, which can be classified as input variables and output variables separately. In traditional DEA models, we needn´t do any transformation for input and output variables when we apply DEA into evaluation procedure because the linear programming can allocate optimal weights for different variables based on the principle of the lower inputs and higher outputs the better. However, traditional DEA models can obtain weak evaluation results that most of DMUs are efficient when each DMU has large amount input variables and output variables. To overcome the shortcoming of DEA models, we proposed an integrated DEA model, denoted as PCA-DEA, in this paper. By using CPA arithmetic, the original data for each DMU can be transferred into uncorrelated data. Then, CPA-DEA models have stronger evaluation capability than traditional DEA models.