Python: Pandas Dataframe AttributeError: 'numpy.ndarray' object has no attribute 'fillna'

Question

Since I am creating a dataframe, I don't understand why I am getting an array error.

M2 = df.groupby(['song_id', 'user_id']).rating.mean().unstack()
M2 = np.maximum(-1, (M - 3).fillna(0) / 2.)  # scale to -1..+1  (treat "0" scores as "1" scores)
M2.head(2)

AttributeError: 'numpy.ndarray' object has no attribute 'fillna'

Solution 1

(M - 3) is getting interpreted as a numpy.ndarray. This implies that M is defined somewhere as a numpy.ndarray. Test it out by running:

print type(M)

Solution 2

You are calling the .fillna() method on a numpy array. And numpy arrays don't have that method defined.

You can probably convert the numpy array to a pandas.DataFrame and then apply the .fillna() method.

Solution 3

Your code is not complete at the moment, so it is hard to pin point why M is causing an error. There could be a couple reasons:

1. You have a typo and (M - 3) should be (M2 - 3)

   M2 = df.groupby(['song_id', 'user_id']).rating.mean().unstack()
   M2 = np.maximum(-1, (M2 - 3).fillna(0) / 2.)  # scale to -1..+1  (treat "0" scores as "1" scores)
   M2.head(2)
   

2. You need to define/convert M as pandas.DataFrame somewhere else in your code

   # With out seeing this part of the code, no one can really help you
   M = pd.DataFrame(...)
   # ...
   # ...
   M2 = df.groupby(['song_id', 'user_id']).rating.mean().unstack()
   M2 = np.maximum(-1, (M - 3).fillna(0) / 2.)  # scale to -1..+1  (treat "0" scores as "1" scores)
   M2.head(2)
   

3. You could convert it to a pandas.DataFrame right before you use it.

   M2 = df.groupby(['song_id', 'user_id']).rating.mean().unstack()
   M2 = np.maximum(-1, (pd.DataFrame(M) - 3).fillna(0) / 2.)  # scale to -1..+1  (treat "0" scores as "1" scores)
   M2.head(2)
   

Comments

[emgf_co]

Yes. I used : y = [np.nan, np.nan, 2, 3, 4, 5, np.nan] y_temp = pd.DataFrame(y).fillna(1E-8)[0].values