SEGMENTER : SOFTWARE FOR SEGMENTATION OF LONG TIME SERIES

This application segments time series applying least squares by constant or linear regression.

 

 

AGREEMENT

-          Use this application for only academic purposes.

-          Never distribute any part of this software.

-          Refer the following papers in your papers:

[1] Aksoy H, Unal NE, Gedikli A, Letter to the editor, Stochastic Environmental Research and Risk Assessment, 21(4): 447-449, 2007.

[2] Gedikli A, Aksoy H, Unal NE, Segmentation algorithm for long time series analysis, Stochastic Environmental Research and Risk Assessment, 22(3): 291-302, 2008.

[3] Aksoy H, Gedikli A, Unal NE, Kehagias A, Fast segmentation algorithms for long hydrometeorological time series, Hydrol. Process. 22: 4600–4608, 2008.

[4] Gedikli A, Aksoy H, Unal NE, Kehagias A, Modified dynamic programming approach for offline segmentation of long hydrometeorological time series, [online], Stochastic Environmental Research and Risk Assessment.

[5] Gedikli A, Aksoy H, Unal NE, AUG-Segmenter: A user-friendly tool for segmentation of long time series, Journal of Hydroinformatics, 12(3): 318-328, 2010.

 

 

Contact:

Assoc.Prof.Dr. Abdullah Gedikli

Istanbul Technical University

Civil Engineering Faculty

34469 Maslak, Istanbul, TURKEY

gedikliab@itu.edu.tr

 

 

VERSION 2.1

Requirements:

-          Windows 32/64 bytes operating system.

-          Microsoft .NET Framework 3.5. If not yet, install it.

 

Click on an image to download application and extract it to an empty directory. No installation required, just run Segmentation.exe.

Main Window

32 bytes                                                         64 bytes

 

Options

 

[Open] Single file

Open a single file, Sample Data in which decimal grouping symbol should be used for neither floating point nor integer numbers.

After pressing [Open] in Open window,

 

[Execute]

The smallest error, which belongs to a segment, greater than zero is reported in error threshold box.

 

[Test] and [Style for output]

Selecting Scheffe test and Numbered style for output:

 

[Plot]

Time series and Segments are plotted in plot pane.

[Plot][Dominants]

Shows dominant change points by vertical red lines.

[Plot][Stamp]

Stamps segments. Blue belongs to Stamped segments (2nd order)

[Plot][UnStamp]

Removes “Stamp”

[Plot][Stamped]

Plots the “Stamped” segments

[Plot][Extract]

Copies the shown segmentation as a new Time Series into clipboard in ASCII text code.

[Plot][Format]

Title : Title of figure as centered at top.

X axis : X axis name

Y axis : Y axis name

Legend : Legend for series

Symbols : Marks data points with circles

 

[Error threshold] box

Also, you can neglect the segments with the error less than an error threshold, say 10.

Here are the results which are the same for error threshold =10 for this example!

[Load] and [Save]

You can save your execution to investigate the segmentations later. Download the executed file ts1a.txt.seg into a directory and press [Load] button then select it to see how it works.

[Open] Multiple file

When opening files, select multiple data files:

Then press [Open]. After execution is completed see the directory including selected files:

Change points of segments are reported in the files, which their names includes “-OUT” . Statistical report is given in “Summary.txt”.

Matlab Interface

Will be available soon.

 

 

PREVIOUS VERSIONS

Version 1.1

 

Console application (by c++)