Research On Purchasing Decision Based On Data Mining

Purchase order management and stock control are two components important of the Enterprise Resource Planning (ERP). In the enterprises with manufacturing mode Make-to-Order (MTO), the accuracy of purchasing quantity (i.e. the accuracy of scrap ratio of materials) influences the stock states, and supplier selection has important position in the purchase order management. In order to improve the accuracy of purchasing quantity and to select supplier optimally, database-supported knowledge acquisition and data mining technologies are usually used to capture objective information and knowledge from mass information. In this dissertation, two advanced theories of data mining—decision tree and Artificial Neural Networks(ANN) are applied to confirm utilization ratio of materials and select suppliers respectively.Firstly, scrap ratio of materials and supplier selection are analyzed deeply and their importance to ERP is described. The theory of decision tree and ANN are explained, and then the application of the two theories is developed.To apply the decision tree theory to scrap ratio of materials, information gain determinant method is used to establish the target attributes, to standardize the decision attributes, to compute the information gain, to prune the tree, and to test the tree, and so on.Similarly, to apply the ANN to supplier selection, corresponding evaluation system for different type of suppliers is designed, the ANN's structure is designed, and then, the ANN model is trained and tested.Finally, the information system is developed as practical research. Oracle8i and MATLAB5.2 are chosen as the tools to develop the system. The results shown that the application research has important practical value to Chinese enterprises.