臺灣大學: 資訊管理學研究所陳靜枝蕭缽Hsiao, BoBoHsiao2013-03-222018-06-292013-03-222018-06-292012http://ntur.lib.ntu.edu.tw//handle/246246/251101雖然網路式資料包絡分析(Network DEA)可以容易描述各階段間之網絡關係，然而對於網路 DEA 具有輸出影響輸入關係卻無法提供任何資訊。 實務上，某些輸入是無法分割於網絡系統的對應輸出(也就是說，某些輸入可能會結合對應的輸出)且可能會影響輸入於特定時間。 本研究提出一種靜態與動態之回餽式的網路結構透過展示其回饋特性與理論介紹引入網路資料包絡分析系統。藉由台灣商業銀行資訊具有回饋系統為例說明並展示研究成果。更進一步，透過一個複雜的效率系統藉由串連方式、並聯方式與本研究提出回饋方式之分析技術說明網路效率拓樸與評估技術。This dissertation proposes a static/dynamic feedback structure for analyzing network efficiency using data envelopment analysis (DEA). The network DEA approach can easily explain the relationship among processes; however network DEA does not give any information about how the outputs of a network system may impact its inputs. In practice, certain inputs are not separable from their corresponding outputs (i.e., certain inputs may combine inputs with the corresponding outputs), and this may influence inputs during the same period. This dissertation demonstrates the characteristics of a feedback system and the theoretical reasoning behind it. We formulate feedback models using data from the interface between the process boundary and the system boundary, which contains one or more process boundaries (i.e., this interface is a point of interaction between the processes or between the system and a given process). We use these models to evaluate efficiency scores employing a banking system example with a feedback effect. Results from our feedback analysis provide more realistic efficiency scores under the feedback effect characteristics. Moreover, we also deconstruct a complicated network into three pattern types and evaluate their overall efficiency scores. After considering the feedback structure of DEA, a complex network DEA problem can be easily composed by the three types of patterns.1087255 bytesapplication/pdfen-US資料包絡分析 (DEA)回饋系統效率網絡拓樸回饋程序回饋效應Data Envelopment Analysis (DEA)Feedback SystemNetwork TopologyFeedback ProcessFeedback Effectsthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/251101/1/ntu-101-D96725002-1.pdf