matrixprofile-ts is a Python 2 and 3 library for evaluating time series data using the Matrix Profile algorithms developed by the Keogh and Mueen research groups at UC-Riverside and the University of New Mexico. Current implementations include MASS, STMP, STAMP, STAMPI and STOMP.
Read the Target blog post here.
Further academic description can be found here.
Major releases of matrixprofile-ts are available on the Python Package Index:
pip install matrixprofile-ts
Details about each release can be found here.
>>> from matrixprofile import *
>>> import numpy as np
>>> a = np.array([0.0,1.0,1.0,0.0,0.0,1.0,1.0,0.0,0.0,1.0,1.0,0.0])
>>> matrixProfile.stomp(a,4)
(array([0., 0., 0., 0., 0., 0., 0., 0., 0.]), array([4., 5., 6., 7., 0., 1., 2., 3., 0.]))
Note that STOMP is highly recommended for calculating the Matrix Profile due to its speed.
Jupyter notebooks containing various examples of how to use matrixprofile-ts can be found under docs/examples.
As a basic introduction, we can take a synthetic signal and use STOMP to calculate the corresponding Matrix Profile (this is the same synthetic signal as in the Golang Matrix Profile library). Code for this example can be found here
There are several items of note:
-
The Matrix Profile value jumps at each phase change. High Matrix Profile values are associated with "discords": time series behavior that hasn't been observed before.
-
Repeated patterns in the data (or "motifs") lead to low Matrix Profile values.
We can introduce an anomaly to the end of the time series and use STAMPI to detect it
The Matrix Profile has spiked in value, highlighting the (potential) presence of a new behavior. Note that Matrix Profile anomaly detection capabilities will depend on the nature of the data, as well as the selected subquery length parameter. Like all good algorithms, it's important to try out different parameter values.
- Andrew Van Benschoten ([email protected])
-
Chin-Chia Michael Yeh, Yan Zhu, Liudmila Ulanova, Nurjahan Begum, Yifei Ding, Hoang Anh Dau, Diego Furtado Silva, Abdullah Mueen, Eamonn Keogh (2016). Matrix Profile I: All Pairs Similarity Joins for Time Series: A Unifying View that Includes Motifs, Discords and Shapelets. IEEE ICDM 2016
-
Matrix Profile II: Exploiting a Novel Algorithm and GPUs to break the one Hundred Million Barrier for Time Series Motifs and Joins. Yan Zhu, Zachary Zimmerman, Nader Shakibay Senobari, Chin-Chia Michael Yeh, Gareth Funning, Abdullah Mueen, Philip Berisk and Eamonn Keogh (2016). EEE ICDM 2016
-
Matrix Profile V: A Generic Technique to Incorporate Domain Knowledge into Motif Discovery. Hoang Anh Dau and Eamonn Keogh. KDD'17, Halifax, Canada.