Abstract: A method separates multivariate data points in lower dimensional space, where each data point has been classified into one of a plurality of data clusters including at least a first data cluster and a second data cluster. The method includes the step of acquiring an ND-to-3D transformation matrix for transforming the plurality of multivariate data points to a plurality of three-dimensional data points. The method preferably includes the sub-step of performing a center of mass (COM) separation of the clusters to acquire a COM transformation matrix, where the COM transformation matrix is the ND-to-3D transformation matrix. The method also includes the step of performing a receiver-operator characteristic curve (ROC) separation to acquire an ROC transformation matrix for transforming the plurality of three-dimensional data points to a plurality of data points in a dimension lower than 3D and preferably a re-optimized COM transformation matrix.

Abstract: A method separates multivariate data points in lower dimensional space, where each data point has been classified into one of a plurality of data clusters including at least a first data cluster and a second data cluster. The method includes the step of acquiring an ND-to-3D transformation matrix for transforming the plurality of multivariate data points to a plurality of three-dimensional data points. The method preferably includes the sub-step of performing a center of mass (COM) separation of the clusters to acquire a COM transformation matrix, where the COM transformation matrix is the ND-to-3D transformation matrix. The method also includes the step of performing a receiver-operator characteristic curve (ROC) separation to acquire an ROC transformation matrix for transforming the plurality of three-dimensional data points to a plurality of data points in a dimension lower than 3D and preferably a re-optimized COM transformation matrix.