在分析经典聚类判别分析方法实质的基础上,提出了一种新的聚类判别分析框架,改进了一种基于样本指标值频度计算的两总体判别分析算法,提高了在对所有参与建立判别模型的样本进行判别时的计算速度;给出了建立在此改进判别分析算法基础上的一种动态聚类判别分析算法的设计,并实现了所有算法。进行相应的实证研究,结果表明以此聚类判别分析框架对给定样本集合进行分析,可以迅速得到多个合理的聚类结果以及对聚类结果的清晰解释,既可以对已有的聚类结果进行验证,又可以进行数据的探索性分析。
This paper gave the essential of classical clustering and discriminating analysis. Improved the efficiency of a two- group discriminating analysis method. Then gave the design of a dynamic clustering analysis method based on this discriminating analysis method, and related empirical studies were also done. The outcome shows that this new class of analysis frame can get more reasonable clustering results with clear explanations, and it not only can validate the known clustering results but also can be used for explorative analysis.