Evaluating Distance Measures and Times Series Clustering for Temporal Patterns Retrieval

This paper presents a new method dealing with similarity search and retrieval of temporal motifs from time series data. The suggested approach firstly creates an index over important time series subsequences, using subdimensional clustering. Then, during the querying process, rather than scanning the whole database for extracting relevant answers for a given query, our method traverses the index represented as centroids of the generated clusters, and search for similar subsequences to the query. Finally, relevant temporal associations can be found between the returned motifs using Formal Concept Analysis and Allen's relations.