Description not matching function name

[emanuelef]

Hi,
just noticed a possible mismatch in the documentation, is it possible that the descriptions for
percentage_of_reoccurring_datapoints_to_all_datapoints and
percentage_of_reoccurring_values_to_all_values
have been swapped ?

tsfresh/tsfresh/feature_extraction/feature_calculators.py

Line 866 in f8a952e

def percentage_of_reoccurring_datapoints_to_all_datapoints(x):

tsfresh/tsfresh/feature_extraction/feature_calculators.py

Line 894 in f8a952e

def percentage_of_reoccurring_values_to_all_values(x):

[nils-braun]

Hi @emanuelef !
Thanks for the issue :-) You are right, there is something unclear there.
In my opinion, I think the descriptions match the code:

   ...
        len(different values occurring more than once) / len(different values)
    This means the percentage is normalized to the number of unique values,
    ...
np.sum(counts > 1) / float(counts.shape[0])

and

    ...
        # of data points occurring more than once / # of all data points
    This means the ratio is normalized to the number of data points in the time series
    ...
reoccuring_values / x.size

However, one could argue that the function names themselves are swapped!
Would you like to do a pull request to change that?

[emanuelef]

Sure, I guess also tests needs to be fixed then:

tsfresh/tests/units/feature_extraction/test_feature_calculations.py

Line 518 in f8a952e

self.assertAlmostEqualOnAllArrayTypes(percentage_of_reoccurring_datapoints_to_all_datapoints, [1, 1, 2, 3, 4],

I also wanted to ask why one has two decorators:
@set_property("fctype", "simple")
@set_property("input", "pd.Series")
but the other one misses "input"

Thanks

[emanuelef]

Another question is about the name of the test for percentage_of_reoccurring_values_to_all_values
which is currently test_ratio_of_doubled_values

tsfresh/tests/units/feature_extraction/test_feature_calculations.py

Line 526 in f8a952e

def test_ratio_of_doubled_values(self):

[emanuelef]

Last question is about creating the PR, I cannot create a branch here, ok to create the PR from a fork ?

[nils-braun]

Let me try to answer:

• Yes, forking is the usual way (not only for tsfresh, but basically for every project on GitHub)
• Feel free to change every naming that makes sense to you! We can discuss the details on the pull request :-)
• The input parameter controls the type conversion. I need to look up the details, but it looks like it is not needed. You can leave it as it is for now

[nils-braun]

Sorry quick update: the input parameter is fine. The second function uses the "value_counts" method of the pandas series whereas the first one uses numpy functionality.

[nils-braun]

Fixed in referencing PR