Design And Implementation Of An ERP Data Mining System For The Selection Of Supply Chain Partnership

Traditional ERP (Enterprise Resource Planning) has obvious disadvantages in decision-making and forecast in the following aspects:1. It mainly aims at transactional processing of manufacture and management and optimizing the resources allocation and usage in human, finance and material; therefore, it can not utilize the large amount of data accumulated during the operation;2. The administrators make decisions by using simulative data in traditional ERP, which depends heavily on their experiences; thus it can't respond to the changes of market precisely and rapidly and can't meet the needs of large enterprises.Our example is from the ERP system of Jinda Textile Co., Ltd (Jinda ERP). We designs and implements a data mining system integrating data warehouse and OLAP (On-Line Analytical Processing) technology based on Jinda ERP. It can provide decision-making for the enterprises by making full use of the data to mine the optimal supply chain partnership.The main contributions of this paper are as follows:1. A data mining system based on Jinda ERP is developed to integrate processing flow, decision-making flow and forecast flow, which extends the functions of the traditional ERP;2. Nearest Neighbor-Clustering (NNC) algorithm of radial basis function neural network (RBFNN) is introduced to the selection of supply chain partnership in order to solve the problems of "high rate of delayed delivery" and "high rate of shutdown for material"; thus the decision-making which required several days even several weeks before can be finished within several hours;3. The NNC algorithm is improved to better process the mixed data including numeric data and categorical data by using the idea of how to process categorical data in K-prototype algorithm.