This is a collection of 16 time series features contained in the hctsa toolbox coded in C. Features were selected by their classification performance across a collection of mouse fMRI time-series classification problems. (according to the op_importance pipeline)
For information on the full set of over 7000 features, see the following (open) publications:
- B.D. Fulcher and N.S. Jones. hctsa: A computational framework for automated time-series phenotyping using massive feature extraction. Cell Systems 5, 527 (2017).
- B.D. Fulcher, M.A. Little, N.S. Jones Highly comparative time-series analysis: the empirical structure of time series and their methods. J. Roy. Soc. Interface 10, 83 (2013).
Install GSL library using brew (for Mac Users):
$ brew install gsl
If there is a permission denied error, try this and install again:
$ sudo chown -R $(whoami) $(brew --prefix)/*
Check if it is installed correctly with this command (it should print the path of the library)
$ gsl-config --cflags --libs-without-cblas
In Linux, GSL can be simply installed via
$ sudo apt-get install libgsl-dev
The features are efficiently implemented in C and it can also be used in Matlab and Python. Currently it has been tested only on OS X and Linux platforms.
To compile, simply run the makefile inside the 'C' folder using
$ make
Then run the executable file (make sure you have provided the correct csv or text file name containing the time series data)
$ ./run_feat <infile> <outfile>
The outfile is optional. If not provided then it will print in stdout.
Each line of the output corresponds to a feature with the following comma-separated entries: feature value, name and execution time
Go to the ‘Matlab' directory and run mexAll command in Matlab application
mexAll
This will create a bunch of Matlab executable files with name catchaMouse16_<feature_name>.
Now you can test it by run the catchaMouse16_all script
ts_data = randn(100,1) % column vector
catchaMouse16_all(ts_data)
This will return a feature vector.
Or you can also call the features individually, for example “AC_nl_035” :
catchaMouse16_AC_nl_035(ts_data)
The wrapper needs to be build with linked C library before being importable into some python code. Installation procedure for Python 2 and Python 3 are given below. (If there is a permission denial try with 'sudo')
Copy the following command and run it inside the 'Python' directory
$ python setup.py build
$ python setup.py install
To test this module execute the test python script:
$ python test_features.py
Copy the following command and run it inside the 'Python' directory
$ python3 setup_P3.py build
$ python3 setup_P3.py install
To test this module execute the test python script:
$ python3 test_features.py
Now you can import catchaMouse16 in your code and call the individual features
import catchaMouse16
ts_data = np.rand(100,1)
catchaMouse16.SC_FluctAnal_2_dfa_50_2_logi_r2_se2(ts_data)
Or get the feature vector using
from catchaMouse16 import catchaMouse16_all
features = catchaMouse16_all(ts_data)
Please feel free to ask any query in the Issue section. Any suggestions or improvements are welcomed through Pull Requests.