Research On Supplier Evaluation System Based On Decision Tree Algorithm

The supply chain relationship between the supplier companies and customer enterprises is one of the popular management researches, however, the technical support of supply chain management is comparatively weak, which need us further research to find the more effective, harmonious, and integral supply value-added chain. The supplier evaluation is an important task for the improvement of the strategic cooperation's quality between enterprises. Decision tree classification method based on data mining theory is a relatively mature technology, which is used to finding the useful and easy-understood classify knowledge that hidden in business data.This paper mainly include: the research background and significance of the supplier evaluation, and choose the decision tree algorithm as the theoretical basis of supplier evaluation, and more detailed introduced the principle and implementation process of the decision tree algorithm. Based on the limitations of the traditional decision tree classification algorithm, this paper focuses on the improved method of the decision tree algorithm, such as: co-ordinate analysis the sample data and expertise based on information entropy theory, and find the decision attribute value; applied the Minimum Description Path (MDLP) method in delineation of the decision tree; applied the rough set theory in sample data's attribute reduction of the decision tree classification algorithm; and given a decision classification method based on rough set attribute dependence, and applied to supplier evaluation analysis.Used the electronic order transaction data based on RosettaNet protocol as the supplier assessment data extraction sources, make the supplier evaluation method based on the decision tree applied in e-business procurement transactions'supplier evaluation of domestic telecommunications operators and industry, the results show that the decision tree classification method is more concise under the premise that ensuring the accuracy of sort results.