Time Series Clustering Based on I-k-Means and Multi-Resolution PLA Transform
Vuong Ba Thinh, Duong Tuan Anh · 2012
In this paper, we introduce an approach using I-k-Means algorithm combined with kd-tree for clustering of time series data transformed by the multiresolution dimensionality reduction method, MPLA. Taking advantage of the multiresolution property of MPLA representation, we can use an anytime clustering algorithm such as the I-k-Means, a popular partitioning clustering algorithm for time series. Our approach also uses kd-tree to resolve the dilemma associated with the choices of initial centroids and significantly improve the execution time and clustering quality. Our experiments show that our approach performs better than k-means and classical I-k-Means in terms of clustering quality and running time.