A Proposed Method for Assessing Cluster Heterogeneity
share
Assessing how adequate clusters fit a dataset and finding an optimum number of clusters is a difficult process. A membership matrix and the degree of membership matrix is suggested to determine the homogeneity of a cluster fit. Maximisation of the ratio of the overall degree of membership at cluster number lag 1 is also suggested as a method to optimise the number of clusters in a dataset. A threshold factor upon the degree of membership is also suggested for homogeneous clusters. Cluster simulations were given to compare how well the proposed method compares against established methods. This method may be applied to the output of both hierarchical and k-means clustering.
READ FULL TEXT
