This extension enables user to apply Shapelet analysis on their time series data.

For more information see X. Renard, M. Rifqi, G. Fricout, M. Detyniecki : EAST representation: fast discovery of discriminant temporal patterns from time series, ECML/PKDD Workshop on Advanced Analytics and Learning on Temporal Data, Riva Del Garda, Italy (2016)

The basic idea of Shapelets is, that from a collection of batches of time series, subsequences are randomly drawn. These Shapelet candidates are than used to perform a feature transformation on different time series. The Shapelet is compared with the time series of the new data and the minimal distance of the Shapelet candidate and the time series is calculated. If the minimal distances for many batches are small, the Shapelet can be considered to occur often in the time series and is representable as a subsequence.
The distances can also be different for different kind of batches (different labeled batches). So the calculated features may be suitable for a classification of the batches.

Please have a look into the tutorial processes of the operators to get an impression how to configure them.
The extension contains the following operators:

• Create Searchspace:
  • used to draw the Shapelet candidates from a collection of batches. Delivers a Shapelet Model.
• Shapelet Transformation:
  • used to calculate the distance features by applying the Shapelet model on a collection of batches.
• Shapelet Model to ExampleSet:
  • used to convert a Shapelet Model to an ExampleSet, with the attributes containing the values of the Shapelets.
• Select Shapelets by Weight:
  • used to select only the most relevant Shapelets from a Shapelet Model.

Version 0.1.0 (2019-02-01)

• Initial release
• Create Searchspace Operator
• Shapelet Transformation Operator
• Shapelet Model to ExampleSet Operator
• Select Shapelets by Weight Operator