tk.augment_fourier()

[mdancho84]

Implement tk.augment_fourier() similar to how R timetk tk_augment_fourier() and vec_fourier work:

• date_to_seq_scale_factor: https://github.com/business-science/timetk/blob/b902278c4a7596846594502b4a21595f5018c84d/R/vec-fourier.R#L182
• calc_fourier: https://github.com/business-science/timetk/blob/b902278c4a7596846594502b4a21595f5018c84d/R/vec-fourier.R#L166
• fourier_vec: https://github.com/business-science/timetk/blob/b902278c4a7596846594502b4a21595f5018c84d/R/vec-fourier.R#L134
• augment_fourier: https://github.com/business-science/timetk/blob/b902278c4a7596846594502b4a21595f5018c84d/R/augment-tk_augment_fourier.R#L89

Lead: Justin Kurland

#2

[GTimothee]

I can try to help on this one :)

[mdancho84]

That would be great. The challenge is to get the peaks of the sin/cos to match the period or periods given. I want the functionality to be very similar to R timetk augment_fourier which depends on fourier_vec.

[JustinKurland]

Happy for @GTimothee to give this a go as I have been tackling other tasks presently.

[GTimothee]

OK I will do my best, let's start with this one :) I contacted JustinKurland privately and he kindly offered his help :)

[mdancho84]

Sounds great!!

[GTimothee]

Just to let you know I think I found a bug in the original implementatio. I found that this line

pytimetk/src/pytimetk/core/fourier.py

Line 79 in 8b5014b

data['radians'] = 2 * np.pi * (data[date_column] - min_date).dt.total_seconds() / (24 * 3600)

should be before data.copy() or it will break here:

pytimetk/src/pytimetk/core/fourier.py

Line 86 in 8b5014b

np.cos(freq * df['radians']) if order % 2 == 0 else np.sin(freq * df['radians'])

(key error 'radians' in df['radians'])

Should I push a fix in a separate issue or should we ignore it and just wait for the new version of the function ?

[mdancho84]

Just push the fix, and we will get it updated. This way if we modify the original function, then the corrected version has the fix already in there.

[mdancho84]

@GTimothee Just a heads up - I've refactored this and the rest of the feature engineering functions to a new submodule called feature_engineering. Please do a git pull.

[GTimothee]

Sorry I did not see your message ! Will do

[mdancho84]

Done: Here are the docs: https://business-science.github.io/pytimetk/reference/augment_fourier.html#pytimetk.augment_fourier