mbclust

This function performs the model-based clustering: agglomerative clustering, EM, and calculating the BICS. It only does the 9 basic models where we have closed-form updates for the covariance matrices. See the help file on mbcfinmix for more informaiton on the model types.It returns the BESTMODEL based on the highest BIC.

Input Arguments:

DATA is a matrix of observations, one on each row.

MAXCLUS is the maximum number of clusters to look for.

Output Arguments:

• BICS is a matrix of BIC values. The row corresponds to the model and the column represents the number of clusters. You can plot these values using plotbic.
• BESTMODEL is a structure containing the best model - the one with the highest BIC value.
• ALLMODELS is a structure containing the results for ALL models.
• Z contains the cluster structure from the agglomerative model-based clustering. Can be used in dendrogram and rectplot.
• CLABS is the cluster labels according to the BESTMODEL.

The output variable ALLMODELS is a structure containing all of the models. For example, ALLMODElS(I) indicates the I-th model type (1-9) and CLUS(J) indicates the model for J clusters.

Synopsis

[BICS,BESTMODEL,ALLMODELS,Z,CLABS] = MBCLUST(DATA,MAXCLUS);

References:

Model-Based Clustering Toolbox documentation.

Model-Based Clustering homepage.