论文部分内容阅读
本文讨论 T-平方统计在模式类可聚性检验中的应用问题。实践证明,在应用聚类算法之前确定模式集合是否存在聚类结构,可以帮助我们避免对聚类算法的不适当应用,从而有效化聚类结果。Monte-Carlo 实验和对 IRIS 等数据的实际检测指出,T-平方统计结合半数框架应用,对良好聚类的模式集合有较强的检验功效。
This article discusses the application of T-squared statistics in pattern class coherency testing. Practice has proved that before the application of clustering algorithm to determine whether there is clustering pattern collection can help us avoid the improper application of clustering algorithm to effectively cluster results. The Monte-Carlo experiment and the actual detection of data such as IRIS indicate that the combination of T-squared statistics and half-frame application has a strong test effect on the well-clustered pattern set.