Empirical Research on Classifying Optimization Based on Mahalanobis Distance
Ruoyu Lu · Ruan kexue · 2007
This paper chooses some weakness of clustering measurements and methods,adopts the Mahalanobis distance to measure the sample data's distance,puts forward the stability and sensitivity principle of clustering algorithm considering the statistics theory and management practice.Firstly it selects the ward method to perform clustering and gets the possible values of the numbers of clusters through Dendrogram,finally determines the optimized numbers of clusters through solving the maximum value of Pseudo-F statistic.