An approach to linguistic summarization based on comparison among multiple time-series data

This paper proposes a method of linguistic summarization of the relation among multiple time-series data by comparing them. The relation among the data is found by correlation coefficient and then it is categorized into main three relations: (i) similar trends, (ii) symmetrical trends, and (iii) non-correlation. Symbolic Aggregate approximation (SAX) is applied to the data categorized into these three types for coding numerical data, and then significant points of two time-series data are extracted by our modified edit distance.