使用隨機森林及多準則決策方法改善綠色供應商績效

Using random forests and multi-criteria decision method for improving green supplier performance

10.6495/CIMBM.201806.0034

倪薇芳(Wei-Fang, NI)；劉建浩(James Liou)；徐瀞禧(Jing-Si, SYU)

績效指標 ； 資料探勘 ； 隨機森林 ； 多準則決策分析 ； SVM ； Data mining ； Green supplier selection ； MCDM

科際整合管理研討會

2018第20屆（2018 / 08 / 01）

486 - 502

繁體中文

全球化的影響下，企業所面臨最重要的挑戰即是供應商管理。此外，綠色環境績效也逐漸成為供應商管理的必要項目。供應商管理要有適當定期的評量以確保能達到各種品質系統績效評估的指標，本研究為針對合格供應商以持續合作為前提，專注於品質系統的績效指標進行客觀篩選與供應商的績效改善。使用隨機森林(Random Forest, RF)給予指標重要度觀念，篩選出關鍵的品質系統指標。再使用決策實驗室分析法(Decision Making Trial and Evaluation Laboratory, DEMATEL)瞭解指標間的因果關係結構，並結合決策實驗室分析基礎之網路層級分析法(DEMATEL-based Analytic Network Process, DANP)，獲取指標相互影響作用後的權重。最後，使用VIKOR探討各供應商的績效與渴望水準間的差距。結果顯示「最終產品控制」、「品質管理與策劃」、「出貨檢驗」為前三項重視的準則。在供應商績效評估中，也提出改善供應商績效的建議與方向。經由實際運算後，證明此模式之有效性。

Under the influence of globalization, the most important challenge for the enterprise is supplier management, regarded as the best intangible assets. In addition, green environmental performance has become a necessary indicator for supplier management owing to the rising awareness of environment protection. Supplier management should have appropriate evaluation methods and regularly evaluate supplier performance. This study focuses on the quality system performance indicators and the improvement of supplier performance. This study applied the random forest, a data mining method, to extract the essential indicator from the quality auditing data set. Then, the Decision Making Trial and Evaluation Laboratory (DEMATEL) was used to understand the structure of cause-effect relationship between the indicators. By the obtained influence relationship between indicators, this study further utilized the DEMATEL-based Analytic Network Process (DANP) to obtain the weights of the indicators. Furthermore, in order to improve the existing supplier performance, the VIKOR method was used to discuss the performance gaps of each supplier and aspired level.