Abstract

. This paper describes our work on a new automatic indexing technique for large one-dimensional (1D) or time-series data. The principal idea of the proposed time-series data indexing method is to encode the shape of time-series into an alphabet of characters and then to treat them as text. As far as we know this is a novel approach to 1D data indexing. In this paper we report our results in using the proposed indexing method for retrieval of real-life time-series data by its content. 1 Introduction Digitized 1D data, commonly referred to in the literature as time-series data is a sequence of real values and it can represent voice, sensor reading, history of stock prices, etc. Although we propose to investigate indexing of any sequence of real numbers we use time-series to be consistent with the literature. Timeseries data are generated and stored in many applications. Examples include voice, histories of stock prices, histories of product sales, histories of engine testing, sei...

Citations