Problem: The function runs slowly for large input data sets.
For a large input data set, the function runs slowly, spending much time on one step, or it terminates with a failure message. Consult the logs for error messages and troubleshooting information.
Workarounds:
- Improve the execution time of the saxification step, in any of the following ways:
- Decrease the difference between the SaxMinWindowSize and SaxMaxWindowSize argument values.
- Increase the SaxOutputFrequency argument value.
- Decrease the SaxSymbolsPerWindow argument value.
- Decrease the number of masking operations by decreasing the RandomProjections argument value.
- Decrease the number of shapelets in the output table by decreasing the ShapeletCount argument value.
- Increase the number of data points to skip between consecutive time series windows when calculating the distance of a shapelet from a time series by increasing the TimeInterval argument value.
Problem: Classification accuracy is not good enough.
The function completes successfully, but the classification accuracy is low.
Workarounds:
- Improve the accuracy of the saxification step, in any of the following ways:
- Increase the difference between the SaxMinWindowSize and SaxMaxWindowSize argument values.
- Decrease the SaxOutputFrequency argument value.
- Increase the SaxSymbolsPerWindow argument value.
- Increase the number of masking operations by increasing the RandomProjections argument value.
- Increase the number of shapelets in the output table by increasing the ShapeletCount argument value.
- Decrease the number of data points to skip between consecutive time series windows when calculating the distance of a shapelet from a time series by decreasing the TimeInterval argument value.