WaveSim and Adaptive WaveSim Transform for Subsequence Time-Series Clustering
R. Pradeep Kumar, Panduranga Naidu Nagabhushan, Ahlame Chouakria-Douzal · 2006
Recent days advancement in sensor and instrumentation technology has seen a large amount of time series data being recorded in our day-to-day life. Knowledge and data mining research has taken up the responsibility of mining the hidden patterns in these huge collection of time series data during the past decade. In this paper we propose methodologies to extract hidden knowledge in a time series data through an unsupervised approach by using the novel WaveSim transform. This recently introduced transform is a novel perspective of wavelet transform and it is defined by keeping pattern analysis and recognition in mind. Time series data mining has been classified broadly into whole series mining and subsequence series mining. We propose a hierarchical tree based approach for subsequence mining in a time series using a modified WaveSim transform called Adaptive WaveSim transform. The technique has been illustrated through a set of experimentation results which is expected to open up a wide arena for future work.