Is there a way to transform new data after fitting with FAMD?

[kuhanw]

Hello,

I just discovered this package and it seems very interesting. I was wondering is there a way to apply the transform function to new unseen data after calling FAMD fit? Analogous to how PCA works in sklearn.

When I try to do this I get an error:

X)
102 X = self.scaler_.transform(X)
103
--> 104 return pd.DataFrame(data=X.dot(self.V_.T), index=index)
105
106 def row_standard_coordinates(self, X):

ValueError: shapes (2,20) and (49,2) not aligned: 20 (dim 1) != 49 (dim 0)

Basically it looks like it doesn't understand there are a different number of "training examples" as opposed to when the fit occurred.

Cheers,

Kuhan

[MaxHalford]

Hey,

I'm not exactly sure what you're trying to do, could you possible provide a more complete example? Normally if you call the transform method of the FAMD object it should work as long as the number of columns is the same in the new dataset.

[kuhanw]

Hi Max,

In sklearn I could have:

a = [some data]

a0 = a[:n]
a1 = a[n:]
pca = PCA()
pca.fit(a0)
pca.transform(a1).

Basically performing the PCA on part of the data and then using it to transform another unseen portion of data. Is it possible to do this with the MCA or FAMD modules in Prince?

Cheers,

Kuhan

[MaxHalford]

It should be possible, but it looks something is going wrong... I'll look into now.

[kuhanw]

Thanks for the reply. Let us know how it goes!

Cheers,

Kuhan

[MaxHalford]

Well I think I fixed the issue for the FAMD, is that what you had a problem with? If not can you copy/paste the entire code you're having a problem with?

[kuhanw]

Hi Max,

No. See below:

X = pd.DataFrame(
data=[
['A', 'A', 'A', 2, 5, 7, 6, 3, 6, 7],
['A', 'A', 'A', 4, 4, 4, 2, 4, 4, 3],
['B', 'A', 'B', 5, 2, 1, 1, 7, 1, 1],
['B', 'A', 'B', 7, 2, 1, 2, 2, 2, 2],
['B', 'B', 'B', 3, 5, 6, 5, 2, 6, 6],
['B', 'B', 'A', 3, 5, 4, 5, 1, 7, 5]
],
columns=['E1 fruity', 'E1 woody', 'E1 coffee',
'E2 red fruit', 'E2 roasted', 'E2 vanillin', 'E2 woody',
'E3 fruity', 'E3 butter', 'E3 woody'],
index=['Wine {}'.format(i+1) for i in range(6)]
)

famd.fit(X[:4])

famd.transform(X[4:])

---

ValueError Traceback (most recent call last)
in ()
----> 1 famd.transform(X[4:])

c:\users\wangku\appdata\local\continuum\anaconda3\envs\tensorflow-gpu\lib\site-packages\prince\mfa.py in transform(self, X)
121 def transform(self, X):
122 """Returns the row principal coordinates of a dataset."""
--> 123 return self.row_coordinates(X)
124
125 def _row_coordinates_from_global(self, X_global):

c:\users\wangku\appdata\local\continuum\anaconda3\envs\tensorflow-gpu\lib\site-packages\prince\mfa.py in row_coordinates(self, X)
135 # Prepare input
136 X = self._prepare_input(X)
--> 137 return self._row_coordinates_from_global(self._build_X_global(X))
138
139 def row_contributions(self, X):

c:\users\wangku\appdata\local\continuum\anaconda3\envs\tensorflow-gpu\lib\site-packages\prince\mfa.py in _row_coordinates_from_global(self, X_global)
125 def _row_coordinates_from_global(self, X_global):
126 """Returns the row principal coordinates."""
--> 127 return len(X_global) ** 0.5 * super().row_coordinates(X_global)
128
129 def row_coordinates(self, X):

c:\users\wangku\appdata\local\continuum\anaconda3\envs\tensorflow-gpu\lib\site-packages\prince\pca.py in row_coordinates(self, X)
102 X = self.scaler_.transform(X)
103
--> 104 return pd.DataFrame(data=X.dot(self.V_.T), index=index)
105
106 def row_standard_coordinates(self, X):

ValueError: shapes (2,11) and (12,2) not aligned: 11 (dim 1) != 12 (dim 0)

If you fit on the first three rows rows X[:3] and try to transform the last three X[3:] it will work. Something to do with the shape?

Cheers,

Kuhan

[MaxHalford]

I think I know what this is due to. I'll fix it over the weekend :)

[MaxHalford]

@kuhanw sorry for the delay, can you install the latest code from GitHub and tell me if it works? It should do. You can install it by running pip install git+https://github.com/MaxHalford/Prince.

[kuhanw]

I did some quick testing and it seemed to work. I will try it on some more complicated datasets later and see if it holds. Thanks for looking into this.

Cheers,

Kuhan

[ggjuancamilo]

Hi, i'm getting the same issue

[pauloeddias]

This is also happening to me now with FAMD

[msalem7777]

This issue is back in 0.7.1