This Repository consists of code to run a correlation based feature subset selection technique for multivariate time series data called MSTS
The code runs the data from the UEA multivariate time series archive and comapres against a wrapper approach
There are 4 main functions presented here:
DTW_MSTS.py - Runs the MSTS technique using a 1NN-DTW classifier
Rocket_MSTS.py - Runs the MSTS technique using a Rocket classifier
DTW_knn.py - Runs the knn based MI technique using a DTW classifier
Rocket_knn.py - Runs the knn based MI technique using a Rocket classifier
When using this code, kindly cite the following paper:
Kathirgamanathan, B., Cunningham, P.: Correlation based feature subset selection for multivariate time-series data. arXiv preprint arXiv:2112.03705 (2021) https://arxiv.org/pdf/2112.03705.pdf