Sequence Classification: A Regression Based Generalization of Two-stage Clustering
Nusrat Jahan Farin, Nafees Mansoor, Sifat Momen, Iftekharul Mobin, Nabeel Mohammed
Conference Paper. International Workshop on Computational Intelligence, (IWCI 2016), December 2016, Bangladesh (2016).
Abstract
Two-stage clustering based approaches has been used for sequence classification. However, certain parameters of these process are either hand picked or found through exhaustive searches. In this paper we propose a simple regression based approach, derived from parameter values found in a previous study, which generalizes the method to find values of three different parameters through proposed equations. We tested the applicability of our method on eight UCR sequence datasets and found our method to be comparable with hand picked approaches and better than single clustering based approaches. © 2016 IEEE.
Keywords
cluster number, kmeans++, sliding window size, time-series data