Skip to content
/ ustc Public

Uncertain Shapelet Transform Classification, a shapelet method for uncertain time series classification

Notifications You must be signed in to change notification settings

frankl1/ustc

Repository files navigation

Usage

Check the file u_shapelet.ipynb for tutorial.

Rerun the experiment using the file run_in_parallel.py.

The datasets: uncertain-dataset.tar.gz

Results

  • The accuracy of each considered method and each uncertainty level is here: all_results.csv

Critical difference diagrams of models accuracy rank(lower is better)

Low uncertainty Medium uncertainty High uncertainty
NB ./images/CD_ust_models_nb_01.png ./images/CD_ust_models_nb_06.png ./images/CD_ust_models_nb_16.png
RF ./images/CD_ust_models_rf_01.png ./images/CD_ust_models_rf_06.png ./images/CD_ust_models_rf_16.png
All ./images/CD_ust_models01.png ./images/CD_ust_models06.png ./images/CD_ust_models16.png

Accuracy scatter plots of UST(UED,RF) vs others

  • Low uncertainty

    ./images/scatter_ulevel01.png

  • Medium uncertainty

    ./images/scatter_ulevel16.png

  • High uncertainty

    ./images/scatter_ulevel06.png

Critical difference diagrams of models log loss (lower is better)

Low uncertainty Medium uncertainty High uncertainty
NB ./images/CD_nb_losslog01.png ./images/CD_nb_losslog06.png ./images/CD_nb_losslog16.png
RF ./images/CD_rf_losslog01.png ./images/CD_rf_losslog06.png ./images/CD_rf_losslog16.png
All ./images/CD_losslog01.png ./images/CD_losslog06.png ./images/CD_losslog16.png

Dependencies

  • imbalanced-learn=0.7.0
  • numpy==1.19.5
  • pandas==1.2.0
  • sktime==0.5.1

Cite this work

@INPROCEEDINGS{mbouopda@2020,  
  author={Mbouopda, Michael Franklin and Mephu Nguifo, Engelbert},  
  booktitle={2020 International Conference on Data Mining Workshops (ICDMW)},   
  title={Uncertain Time Series Classification with Shapelet Transform},   
  year={2020},  
  volume={},  
  number={},  
  pages={259-266},  
  doi={10.1109/ICDMW51313.2020.00044}
}