融合粗糙集和模糊聚类的连续数据知识发现

知识自动获取是困扰基于知识的系统普遍推广应用的瓶颈，粗糙集理论是一种从历史数据中发现规则知识的数学工具。该文针对粗糙集方法应用于电厂与电力系统数据挖掘中存在的连续属性离散化问题，提出了基于模糊聚类的离散化方法。采用模糊C平均(FCM)算法离散连续属性，获得各类的聚类中心以及属性值隶属于各聚类中心的隶属度矩阵，得到离散化的数据。将粗糙集方法应用于离散化后的数据挖掘隐含在历史数据中的知识。最后进一步讨论了置信度、支持度等指标对规则的评价方法。给出的汽轮机轴系振动故障诊断规则获取算例验证了整个知识发现方案的可行性。

COMBINING ROUGH SET METHODOLOGY AND FUZZY CLUSTERING FOR KNOWLEDGE DISCOVERY FROM QUANTITATIVE DATA

Knowledge acquisition is the technique bottleneck of application of knowledge-based systems. Rough set is an emerging tool to extract knowledge from a vast volume of data. Knowledge acquisition technique based on rough set is just suitable for discrete data, but most of raw data in power plants are continuous. To deal with the problem fuzzy clustering is introduced to discretize the continuous attributes. Fuzzy c-mean clustering algorithm is used in our work. A clustering center vector and a membership matrix can be gotten with the algorithm. By this way the continuous attributes is discretizated into decision table,which reduction and rule mining technique based on rough set theory can deal with. Some measures, support and confidence, evaluating the rough rules are analyzed in the paper.A demonstration extracting diagnosis rules from fault data proves the data mining solution is feasible.