or

Can time series classifiers benefit from synthetic data augmentation? In the notebooks in this repository, I explore the application of synthetic data augmentation using weighted dynamic time warping barycenter averaging to improving the performance of a 1-NN DTW-based classifer for multivariate time series.

Notebooks:

1. summary.ipynb - An overview of the methodology and results
2. syntheticdatageneration.ipynb - Synthetic data generation
3. testandplotclassificationaccuracy.ipynb - Testing the classification accuracy as a function of the degree of data augmentation
4. rawtimeseriesvisualization.ipynb - Plotting raw time series data
5. 3DtSNEvisualizations.ipynb - Plotting a 3D t-distributed stochastic neighbor embedding of baseline and augmented datasets