A dynamic programming algorithm for optimal clustering multidimensional data with sequential constraint. The algorithm minimizes the sum of squares of within-cluster distances. The sequential constraint allows only subsequent items of the input data to form a cluster. The sequential constraint is typically required in clustering data streams or items with time stamps such as video frames, GPS signals of a vehicle, movement data of a person, e-pen data, etc. The algorithm represents an extension of Ckmeans.1d.dp to multiple dimensional spaces. Similarly to the one-dimensional case, the algorithm guarantees optimality and repeatability of clustering. Method clustering.sc.dp can find the optimal clustering if the number of clusters is known. Otherwise, methods findwithinss.sc.dp and backtracking.sc.dp can be used.
|Author||Tibor Szkaliczki [aut, cre], J. Song [ctb]|
|Date of publication||2015-05-04 09:27:25|
|Maintainer||Tibor Szkaliczki <firstname.lastname@example.org>|
|License||LGPL (>= 3)|
backtracking.sc.dp: Backtracking Clustering for a Specific Cluster Number
clustering.sc.dp: Optimal Clustering Multidimensional Data with Sequential...
findwithinss.sc.dp: Finding Optimal Withinss in Clustering Multidimensional Data...
print.clustering.sc.dp: Print the result returned by calling clustering.sc.dp