基于粗糙集的CBR检索在天气预测中的应用  

Application of CBR retrieval method to weather forecast based on rough set

作　　者：季赛[1] 沈星[1] 沈超[2] 

机构地区：[1]南京航空航天大学智能材料与结构航空科技重点实验室,江苏南京210016 [2]南京信息工程大学计算机科学与技术系,江苏南京210044

出　　处：《计算机工程与设计》2005年第11期2898-2901,共4页Computer Engineering and Design

基　　金：航空科学基金项目(04G52042)

摘　　要：针对灾害性天气预报和警报的迫切需要,研究了从气象数据中挖掘出知识、积累预报员的经验建立范例库,并利用CBR技术实现相似范例的检索,从而提高天气预报的水平。提出了快速检索的思路,先利用粗糙集理论去除冗余的范例决策表的特征属性,再利用BP神经网络来实现相似范例的检索。该检索方法速度快,效率高。给出了该方法在天气情况预测上的实例,验证了其有效性和准确性。To meet the urgent needs of disastrous weather forecast and alarm, the knowledge should be extracted from the raw meteorological data and the experiences of weather forecaster should be accumulated in order to build the cases database, and to achieve the retrieval of similar cases with CBR technology. Retrieval is the key technology in CBR. It imposes a direct effect on the efficiency and quality of CBR.the rough set is used to reduce the features of the case and BP neural network is used to retrieve the similar cases. This method has high retrieval speed, and the instance is made to test the method. Running result shows the validity and veracity of the method.

关 键 词：粗糙集 属性离散化 范例推理 知识表达系统 BP算法 天气预测 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]