Time-Series Similarity Queries Employing a Feature-Based Approach

Abstract

Time-series, or time-sequence, data show the value of a parameter over time. A common query with time-series data is to find all sequences which are similar to a given sequence. The most common technique for evaluating similarity between two sequences involves calculating the Euclidean distance between them. However, many examples can be given where two similar sequences are separated by a large Euclidean distance. In this paper, instead of calculating the Euclidean distance directly between two sequences, the sequences are transformed into a feature vector and the Euclidean distance between the feature vectors is then calculated. Results show that this approach is superior for finding similar sequences. 2.