规则抽取的Fuzzy ARTMAP方法  

Fuzzy ARTMAP Method of Rule Extraction

作　　者：王立斌[1] 杨毓英[1] 史习智[1] 

机构地区：[1]上海交通大学

出　　处：《数据采集与处理》1998年第4期315-319,共5页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金

摘　　要：着重阐述了如何使用有教师监督的自组织神经网络——模糊自适应共振映射网络（ＦｕｚｙＡＲＴＭＡＰ）从例子中抽取知识规则。叙述了规则抽取中的两个细节：网络修剪，即删除那些对网络抽取规则贡献不大的节点及其相连的权值；权值的量化，以使系统最终能翻译成一套可使用的规则。本文对ＦｕｚｚｙＡＲＴＭＡＰ网络作了改进和简化，并用于医学上心电图（ＥＣＧ）信号中室性早博（ＰＶＣ）诊断规则的自动获取，取得了比较满意的结果。A new method of rule extraction is discussed in this paper, which shows how to extract the knowledge, in the form of fuzzy rules, from signals by using the supervised self organizing neural network, i.e. fuzzy adaptive resonance mapping network called fuzzy ARTMAP. Two details in the rule extraction, pruning and quantization, are described. The former deletes those nodes and connecting weights that don′t greatly contribute to the rule extraction and the latter can translate the weights into a set of usable rules. Some improvements and simplification on Fuzzy ARTMAP are made, and a comparison with BP networks is also given, finally some diagnostic rules of PVC from ECG by means of Fuzzy ARTMAP are extracted, which are in accordance with the experience of medical experts.

关 键 词：心电图 室性早搏 规则抽取 模糊识别 ARTMAP法 

分 类 号：R540.41[医药卫生—心血管疾病] TP391.4[医药卫生—内科学]