贝叶斯DLM的销售预测系统的研究与实现

Study and Realization of a Forecasting System for Steel Selling Based on Bayesian Dynamic Model

针对某大型钢铁公司对钢铁销售情况预测的需求,在公司海量的钢材销售数据的基础上建立了预测分析模型,以java作为开发语言,建立起基于oracle的数据仓库,数据挖掘算法采用贝叶斯动态模型(DLM-Dynamic Linear Models)预测算法,通过递归运算,提高运算的精度,最后用图形的仿真来预测钢厂的销售情况,形成了用于数据挖掘的预测分析系统。通过了现场的实际运行,将实际值和预测值进行对比,验证了模型建立的正确性,同时预测值达到了较高的预测精度。对管理者的决策提供了重要的参考价值,实际应用效果良好,可以在钢铁行业内推广应用。

According to steel sale of a large steel company,the paper has built a database by using oracle database based on vast amounts of steel selling data of the steel company,and proposed a forecasting analysis model by means of DLM（Dynamic Linear Models）,then the forecasting analysis system used for data mining has come into being finally.The on-site system running has verified the correctness of the constructed model through comparing real value with predicted value,and the model also reaches a higher forecasting precision.The predicted value has important reference value for managers to make decisions.The system has been tested and proven to be useful in practice,and also can be promoted in the field of steel industry.