All notable changes to this project will be documented in this file.
Migration from GitLab to GitHub.
- Added Author list.
- Added option to plot pattern IDs of the embedding in visualization.
- Added option to pass time stamps to the visualization plots.
- Method to convert timedelta values and datetime indices to a number of observations
- Methods to plot the time series and embedding have been moved to
patsemb.pattern_based_embedding
- Raise exception when the window size used for computing the pattern based embedding is larger than the size of the time series
- Exceptions are changed to ValueError when seemed suitable (in
PatternBasedEmbedder)
First release of PaTSEmb! While our toolbox is still a work in progress,
we believe it is already in a usable stage. Additionally, by publicly releasing
PaTSEmb, we hope to receive feedback from the community! Be sure to check
out the documentation
for additional information!
A summary of the available modules within PaTSEmb is given below:
discretization: a module converting a time series into a symbolic representation.pattern_based_embedding: a module for effectively transforming a time series into the pattern-based embedding.pattern_mining: a module for mining patterns within the symbolic representations.postprocess: a module for postprocessing the constructed pattern-based embedding.visualization: a simple module to visualize the time series and embedding.