protti/featts

FeatTS is a Semi-Supervised Clustering method that leverages features extracted from the raw time series to create clusters that reflect the original time series.

FeatTS

Paper

At this link you can find the paper related at this code: http://openproceedings.org/2021/conf/edbt/p270.pdf

Running

python_version > '3.7'

In the testFeatureExtraction.py we can found the main file where we can set the parameter for launch the code. The dataset used for the test should be inside the DatasetTS folder. Inside this folder, you have to create a folder with the same name of the dataset.

At the end of the computation, a file called experiments.tsv will contain all the results obained on the datasets.

Configuration File

For test some other dataset it's very important to create a .tsv file where the first column will be the class of the time series and then all the points of the latter:

Class	1	2	3	4	5	...
0	2.5	2.8	2.2	2.1	3.8	...
1	10.5	12.1	11.2	10.3	14.8	...
0	1.5	1.9	2.2	2.9	3.3	...
...	...	...	...	...	...	...

There is also the possibility to use the .arff files of the UCR Time Series Dataset. Indeed, adding the dataset_TRAIN.arff and the dataset_TEST.arff files, as shown in the DatasetTS folder, FeatTS automatically generate the .tsv file.