There exist many datasets that can be viewed as multivariate time series, such as the daily high temperature at a locality, sensor recordings in diagnostic systems and scientific data, and music and video recordings. These time series reside in large repositories, and there is often a need to search for particular time series exhibiting certain types of behaviors. Many current approaches to time series search are too slow on large repositories, or constrain the range of possible queries.

The Multivariate Time Series Search Capability indexes and efficiently searches a large collection of multivariate time series to rapidly and accurately respond to queries for potentially complex behaviors. It eliminates the need to wade through large numbers of irrelevant results, or miss highly pertinent ones by combining high precision with high recall.

The user can specify a desired time series over multiple variables and allowable ranges over selected variables. The software quickly returns a list of time series within a large multivariate time series database, within the specified range from the query. The software searches for multivariate time series sub-sequences with arbitrary time shifts between the variables and a guarantee to return all possible matches. Current algorithms only allow some of this functionality.

This work was done by Santanu Das, Nikunj Oza, and Ashok Srivastava of Ames Research Center; Kanishka Bhaduri of Stinger Ghaffarian Technologies; and Qiang Zhu of Mission Critical Technologies. The software is hosted at: http://ti.arc.nasa.gov/opensource/projects/mts-search/. NASA invites companies to inquire about partnering opportunities. Contact the Ames Technology Partnerships Office at 1-855-627-2249 or This email address is being protected from spambots. You need JavaScript enabled to view it.. Refer to ARC-16452-1.