摘要
基于单一知识发现方法的不足,提出了一种获取模糊规则的集成方法。首先用Kohonen网络进行数据量化,然后运用粗集理论产生初始规则,并根据所得的规则建立模糊神经网络模型,优化模糊规则的参数,最后再进一步简化获取模糊规则。通过实例进行系统仿真,结果表明该方法是有效的, 同时为获取模糊规则提供了新的思路。最后与其它方法进行了比较,并总结了该方法的特点。
Considering the insufficency of single KDD method, a new hybrid method to acquire fuzzy rules is presented. Firstly, original data are quantized with kohonen network. Then, original fuzzy rules based on rough sets are produced, a fuzzy neural network model is founded and parameters of fuzzy rules are optimized. Lastly, acquired fuzzy rules are further reduced. An example is used to perform system simulation and conclusion indicates that the method is very effective and provides a new idea for acquiring fuzzy rules. Finally, the method is compared with others and its characteristics are summed up.
出处
《自动化与仪器仪表》
2005年第2期5-7,45,共4页
Automation & Instrumentation
