Abstract: In unsupervised learning algorithm, classification results is often not satisfactory. However in supervised learning algorithm, training samples needed to be chosen by manual, which is sometimes difficult, and moreover whose classification accuracy depend on the training samples. According to these limitations, this paper presents an automated pattern classification method of Self-Organizing Map Neural Network (SOMNN) combining with Generalized Regression Neural Network (GRNN). In new method at first original data points are partitioned by unsupervised SOM, and then from the clustering results, some labelled points nearer to each clustering center are chosen to train supervised GRNN. Then utilizing the decided GRNN model, we reclassify these original data points and gain new clustering results. At last from new clustering results, we choose some labelled points nearer to new clustering center to train and classify again, and so repeat until clustering center no longer changes. Experimental results for both Iris data and Wine data verify the validity of our method.

Key words: SOM, GRNN, Pattern automated Classification, PSO algorithm

摘要： 非监督学习算法的分类精度通常很难令人满意，而监督的学习算法需要人工选取训练样本，这有时很难得到，并且其分类精度直接依赖于所选取的学习样本。针对这些缺陷，本文提出一种非监督自组织神经网络（SOMNN）和监督的广义回归网络（GRNN）结合的全自动模式分类新方法。新方法首先通过SOMNN将原始数据进行自动聚类，再用所得的聚类中心以及中心邻近数据点训练GRNN，然后根据GRNN的分类结果重新计算聚类中心，再根据新的聚类中心和中心邻近点训练GRNN，如此反复，直至得到稳定的中心为止。Iris数据，Wine数据的实验结果都验证了新方法的可行性。

关键词: 自组织神经网络, 广义回归网络, 模式自动分类, 粒子群优化算法