A self-organizing clustering algorithm for functional data

Based on a learning schema, we proposed a self-organizing clustering algorithm for functional data. The proposed algorithm can be used to cluster functional data without any prior knowledge of present clusters in the data set. The resulting clusters represent homogeneity in the input data. The theoretical analysis of the convergence of the proposed algorithm is also given. Comparisons with three well-known clustering algorithms, fuzzy k-means, k-means and Funclust, show that the proposed algorithm outperforms these competitors and provides better corrected classification rate and value of variance ratio criterion.