神经网络在确定关联规则挖掘算法权值中的应用研究  

Study of neural network applied to weighted association rules mining

作　　者：李彤岩[1] 李兴明[1] 

机构地区：[1]电子科技大学宽带光纤传输与通信网技术教育部重点实验室,成都610054

出　　处：《计算机应用研究》2008年第5期1440-1441,1449,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60572091)

摘　　要：提出了运用神经网络确定权值的方法,将网络告警信息的三个主要属性作为神经网络的输入,通过样本的训练来确定神经网络的连接权,从而识别网络告警的权值。这种权值确定法既体现了专家的经验知识,又能够随着网络拓扑的变化更新连接权。建模及仿真结果表明,与其他权值确定方法相比,神经网络方法更加实用和有效。A novel method which used neural network to identify the alarm weight was proposed. The neural network had three inputs with the key elements which reflected the importance of the telecommunication alarm. After the course of sample training, the link weight would be confirmed. The weight of the neural network might reflect the knowledge of the experts and also could be changed automatically with the different items from the inputs. Modeling and simulation study indicates that compared with other methods of measuring alarm weight, the neural network method has more advantages.

关 键 词：加权关联规则 通信网告警相关性分析 神经网络 连接权 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]