應用案例與屬性篩選機制建構廠商分群供應鏈績效指標因果模式之研究

The Study of Introducing Instance and Attribute Filtering Mechanism to Construct Causality Model of Firms Grouping Supply Chain Performance Measurement Metrics

辛憶艷(Yi-Yen Hsin)

指導教授 ： 張淳智

國立臺中科技大學/資訊與流通學院/流通管理系碩士班/碩士(2012年)

摘要

由於台灣製造業體系的改變，有關供應鏈績效評估尚缺乏明確的模式可供參考；加上不同的產業別所組成之供應鏈範疇也有所不同。因此，若能透過專家系統的方式建立適合各產業類型之供應鏈評估模式，將有助於企業進行績效評估以及改善建議。
本研究係基於案例式推論為理論基礎，結合基因演算法與迴歸模式，發展一套供應鏈績效關鍵衡量指標之萃取系統雛型。其中，本研究導入案例與屬性篩選技術，即二維度縮減技術，應用基因演算法強大的演化能力尋找最佳或近似最佳解之分群，減少受雜訊的影響；另外，本研究發展一種新的染色體編碼方式，將分群資訊簡化為bit的階層表現，大幅度降低記憶體佔用空間，縮短求解問題之時間。
再者，為了驗證本研究分群之合理性與各群體之解釋能力，在模式中加入符號判斷以確保合理性，並使用各群體之調整後判定係數作為衡量準則。同時，比較不同模式分群之預測準確性，將集群分析法中的K-means模式在相同資料與分群數目進行求解，其結果遠低於本研究模式的預測水準。
透過本研究建構流程可供業界進行模式建立與績效改善之參考。分群結果將可作為企業評估稽核以及策略發展的依據。

關鍵字

二維度縮減 ； 基因演算法 ； 供應鏈管理 ； 迴歸模式

並列摘要

Due to the change of manufacturing industry framework in Taiwan, there was no accurate model of Supply Chain Performance Measurement for reference yet and the consisted supply chain scope is different in every industry category. Therefore, if the supply chain adopted by Taiwan manufacturing industry can use the supply chain measurement model constructed through the expert system which is suitable for each field of industry, it is helpful for enterprise to do performance measurement as well as providing suggestions for improvement.
This study introduced Case-based Reasoning as the theoretical basis. By utilizing genetic algorithm and regression analysis, a preliminary system model to extract supply chain performance key evaluation index was developed. We introduced instance and attribute filtering technique, i.e. two-dimensional reduction technique, with the help of powerful capability of gene algorithm to search for the optimum or nearly optimum solution to reduce the noise impact. In addition, the study developed a new chromosome encoding method to simplify grouping information into bit level, which drastically reduced memory consumption and shorten the time for solution acquirement.
Furthermore, in order to verify the grouping reasonability and interpretability of each group in this study, we added -sign judgment to ensure the reasonability, and used Adjusted Coefficient of Determination of each group as the evaluation standard. At the same time, forecast accuracy was compared in different model grouping. K-means clustering method was used to derive the solution in the same data and group number and it is found that the forecast result was much poorer than the proposed model in this study.
The procedure created in this research provides a reference to model construction and performance improvement for industry. The grouping result can be the basis for industry audit evaluation and strategy development.

並列關鍵字

Two-dimensional reduction technique ； Genetic algorithm ； Supply chain management ； Regression analysis

參考文獻

Ahn, H., & Kim, K. J. (2009). Global optimization of case-based reasoning for breast cytology diagnosis. Expert Systems with Applications, 36, 724-734.

Ahn, H., Kim, K. J, & Han, I. (2007). A case-based reasoning system with the two-dimensional reduction technique for customer classification. Expert Systems with Applications, 32, 1011-1019.

Cardie, C. (1993). Using decision trees to improve case-based learning. In Proceedings of the Tenth International Conference on Machine Learning, 25-32.

Chen, Y. K., Wang, C. Y., & Feng, Y. Y. (2010). Application of a 3NN+1 based CBR system to segmentation of the notebook computers market. Expert Systems with Applications, 37, 276-281.

Chiu, C. (2002). A case-based customer classification approach for direct marketing. Expert Systems with Applications, 22(2), 163-168.