基于遗传算法的非监督层次分类方法研究

Unsupervised hierarchical classification method based on genetic algorithm

快速准确地确定单个样本的所属类别以及总体样本类别数是解决非监督模式识别的前提,然而它们的确定通常是非常困难的。通过研究基于遗传算法的相似性度量最优分类算法以及最优分类数确定算法,提高非监督识别的准确性,并将所研究的算法应用到飞机识别当中。实验结果表明,本算法可以进行最优分类及分类数的确定。

It is important to decide the class of a single sample and the quantity of classes of overall samples quickly and exactly for solving unsupervised pattern classification,which is usually very difficult to realize. We studied the algorithm of comparability measurement optimization classification based on genetic algorithm and the algorithm for confLrrning the quantity of optimization classifications. Experiment results showed that the algorithm proposed can decide optimization classification and determine the quantity of optimization classification.