基于神经网络规则抽取的产品服务配置规则获取  

Product Service Configuration Rules Acquisition Based on Rule Extraction from Neural Networks

作　　者：沈瑾[1,2] 王丽亚[1] 隆惠君[3] 吴明兴[1] 江志斌[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240 [2]上海电机学院商学院,上海201306 [3]上海交通大学中美物流研究院,上海200030

出　　处：《工业工程与管理》2012年第3期66-73,共8页Industrial Engineering and Management

基　　金：国家自然科学基金重点项目(70932004/G0209)

摘　　要：针对由混合配置前件导致的大量耦合的、相互关联的复杂规则,提出了LC神经网络和RULEX算法联合实施下基于历史数据的个性化产品服务配置规则获取方法,包括基于LC的规则构造和知识发现,基于RULEX的规则抽取,网络模块的划分与提炼,以及配置规则适应度评价。最终通过楼宇控制产品服务说明了该方法具有较好的网络训练效率和规则复杂度。To deal with a large number of coupled,connected complex rules caused by mixed antecedents,acquisition methods for personalized product service configuration rules based on LC neural network and RULEX algorithm is proposed in this paper. First, rule knowledge is discovered by LC from historical data, then extracted by RULEX. Network module partition and analysis of rules fitness are also presented. Finally, the proposed method is validated by services of building control product.

关 键 词：产品服务 配置规则 神经网络规则抽取 LC RULEX 

分 类 号：TP3-05[自动化与计算机技术—计算机科学与技术]