Pandas for Python: Exception: Data must be 1-dimensional
Most sklearn
methods don't care about column names, as they're mainly concerned with the math behind the ML algorithms they implement. You can add column names back onto the OneHotEncoder
output after fit_transform()
, if you can figure out the label encoding ahead of time.
First, grab the column names of your predictors from the original dataset
, excluding the first one (which we reserve for LabelEncoder
):
X_cols = dataset.columns[1:-1]X_cols# Index(['Age', 'Salary'], dtype='object')
Now get the order of the encoded labels. In this particular case, it looks like LabelEncoder()
organizes its integer mapping alphabetically:
labels = labelencoder_X.fit(X[:, 0]).classes_ labels# ['France' 'Germany' 'Spain']
Combine these column names, and then add them to X
when you convert to DataFrame
:
# X gets re-used, so make sure to define encoded_cols after this lineX[:, 0] = labelencoder_X.fit_transform(X[:, 0])encoded_cols = np.append(labels, X_cols)# ...X = onehotencoder.fit_transform(X).toarray()encoded_df = pd.DataFrame(X, columns=encoded_cols)encoded_df France Germany Spain Age Salary0 1.0 0.0 0.0 44.000000 72000.0000001 0.0 0.0 1.0 27.000000 48000.0000002 0.0 1.0 0.0 30.000000 54000.0000003 0.0 0.0 1.0 38.000000 61000.0000004 0.0 1.0 0.0 40.000000 63777.7777785 1.0 0.0 0.0 35.000000 58000.0000006 0.0 0.0 1.0 38.777778 52000.0000007 1.0 0.0 0.0 48.000000 79000.0000008 0.0 1.0 0.0 50.000000 83000.0000009 1.0 0.0 0.0 37.000000 67000.000000
NB: For example data I'm using this dataset, which seems either very similar or identical to the one used by OP. Note how the output is identical to OP's X
matrix.