模糊聚类算法及应用

Fuzzy cluster algorithm and its application

模糊聚类是采用模糊数学方法 ,依据客观事物间的特征、亲疏程度和相似性 ,通过建立模糊相似关系对客观事物进行分类的一门多元技术。其算法主要有传递闭包法、动态直接聚类法和最大树法等 ,其中动态直接聚类法计算量最少。在实际应用中必须经过数据预处理、特别是归一化等处理步骤 ,选取合适的模糊关系建立模糊相似矩阵 ,然后进行聚类和模式识别。通过典型数值算例和油水层识别结果表明 ,其应用效果良好。

Fuzzy cluster includes these algorithms,such as the transmitting closure,dynamic direct cluster,and maximum tree,and the operation times of dynamic direct cluster algorithm is the least. It is subject to the data pre-processing and the normalization procedure. The fuzzy similarty matrix is established by using of the appropriate fuzzy relationship,then the clustering and pattern recognition can be realized. The typical numerical example and oil- water formation experiment show the effect is good in actual application.