关键词: 钢铁企业产成品预报, 神经网络, 企业数据挖掘, SAS

Abstract: Based on the order tracking strategy and the implementing approaches in the Integrated Production/Distribution System in an iron-steel enterprise, the product output rule in cool-rolling product line was explored using the SEMMA process of SAS/Enterprise Miner software. An artificial neural-network model for forecasting product output was built. In this forecasting model, release date and time of order at the last several processes in product line were taken as inputs, and finish date of order was taken as output. A synthetic computation experiment with neural network function of SAS/Enterprise Miner shows that the neural network models are effective approaches for forecasting product output and the precision is up to 91%. The experiment also shows that the activation function type of hidden nodes has definite effect on performance, but the number of nodes in hidden layer has little effect on performance, and data noise has some effect on performance.