Published
Thai Journal of Operations Research
ตัวแบบการเรียนรู้จําแนกประเภทซัพพลายเออร์แบบมีผู้สอนสําหรับปัญหาการประเมินประสิทธิภาพของซัพพลายเออร์ในระบบ SAP ERP
Abstract
The purpose of this research is to develop a supplier classification framework for supplier
evaluation problem for overcoming the uncertain results from embedded linear scoring methods in SAP
ERP. Datasets in this study are Purchase Order (PO) and Purchase Requisition (PR). PO data is used for
modelling supplier classification models, and PR is used as an input data to evaluate the suppliers’
performances before making purchasing contracts. Three evaluation criteria, which are developed by
interviewing with experts are quantity, quality, and delivery commitment. In addition, this research is
developed the data extraction algorithm from SAP ERP system and data transformation in order for designing
suitably datasets for supplier classification models. Nine classification models: Naïve bay, K-Nearest
Neighbors, Support Vector Machine, Logistic Regression, Adaboost, Decision Tree, Random Forest, and
Stacking SVM+DT and Stacking SVM+LR are applied to compare the performances to find the best
classification model. In the analysis schemes, all material groups and each material group as suggested two
levels of analysis are proposed in this study. Results show that Adaboost is the best classification model in
both train and test datasets. Furthermore, Adaboost is also outperform others for both levels of analysis
with accuracy performances for all material group (81.6%) and each material group (67.6), respectively.
(2021). ตัวแบบการเรียนรู้จําแนกประเภทซัพพลายเออร์แบบมีผู้สอนสําหรับปัญหาการประเมินประสิทธิภาพของซัพพลายเออร์ในระบบ SAP ERP. Thai Journal of Operations Research, 9(1), 106-119.