A principal curve-based method for data clustering

In this work a new method for data clustering based on principal curves is presented. Principal curves consist of a nonlinear generalization of Principal Component Analysis and may also be regarded as continuous versions of 1-D self-organizing maps. The proposed method divides the principal curves extracted by the k-segments algorithm into two or more curves, according to the number of clusters defined by the user. Then, the data are grouped according to the short distance from them to the curves. The method was applied to eight databases with different characteristics. The results were compared with the k-means algorithm. The method shown to be suitable for elongated and spherical clusters.