Performance evaluation of a two-stage clustering technique for time-series data
Tomoharu Nakashima, Gerald Schaefer, Youhei Kuroda, Md Atiqur Rahman Ahad · 2016
This paper proposes a two-stage clustering technique for time-series data. The proposed method comprises two clustering procedures. First, the original time-series data are divided into subsequences to produce an initial clustering. The resulting group of clusters can be used as features that represent the time-series data. Then, the time-series data are converted into numerical vectors using the features generated by the first clustering stage. Finally, the converted numerical vectors undergo a second clustering procedure that produces the final clustering results. An extensive series of computational experiments are conducted in order to examine the performance of the proposed method.