Abstract— Time course data may inherit critical temporal ordering in contiguous (i.e., neighboring) time slot. Traditional one-way k-means clustering algorithms handle time points independently, ignoring the internal time locality. Although co-clustering algorithms can discover latent local patterns, the discovered patterns are not necessary to be in a continuous time order. Therefore, this paper targets to extend an existing co-clustering framework to be applicable to time course data so that time-dependent local segment patterns over specific intervals can be captured. While following the general co-clustering framework of the alternating optimization process, the proposed algorithms employ clustering on instance dimension and segmentation on time dimension. Both batch and incremental updates at boundary time points are proposed to search for a sequence of time segments. Eight time course datasets and two specific data normalization schemes are considered in the experimental study. Clustering similarity performance among k-means, one existing co-clustering, and the two proposed clustering segmentation algorithms is compared.
Index Terms— Co-clustering, segmentation, pattern discovery, time-course data.
H. Cho and M. K. An are with the Department of Computer Science, Sam Houston State University, Huntsville, Texas, USA (e-mail: hyukcho@shsu.com, an@shsu.com).
[PDF]
Cite: Hyuk Cho and Min Kyung An, " Co-Clustering-Based Clustering and Segmentation for Pattern Discovery from Time Course Data," International Journal of Information and Electronics Engineering vol. 4, no. 5, pp. 358-364, 2014.
