Python - How to reverse the encoding of data encoded with LabelEncoder after it has been split by train_test_split?

3.3k views Asked by At

I am trying to export an unecoded version of a dataset which was encoded using LabelEncoder (from sklearn.preprocessing, to enable application of machine learning algorithms) and was subsequently split into training and test datasets (with train_test_split).

I want to export the test data to excel but with the original values. The examples that I've found till now, use the inverse_transform method of the LabelEncoder on only one variable. I want to apply it automatically on multiple columns that were encoded in the first place.

Here's an example data:

# data
code = ('A B C D A B C D E F').split()
sp = ('animal bird animal animal animal bird animal animal bird thing').split()
res = ('yes, yes, yes, yes, no, no, yes, no, yes, no').split(", ")
 
data =pd.DataFrame({'code':code, 'sp':sp, 'res':res})
data

Assuming 'res' to be the target variable and 'code' & 'sp' to be the features.

1

There are 1 answers

0
AudioBubble On BEST ANSWER

Here you go:

# data
code = ('A B C D A B C D E F').split()
sp = ('animal bird animal animal animal bird animal animal bird thing').split()
res = ('yes, yes, yes, yes, no, no, yes, no, yes, no').split(", ")
 
data = pd.DataFrame({'code':code, 'sp':sp, 'res':res})
data

enter image description here

# creating LabelEncoder object
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()

# encoding
dfe = pd.DataFrame()    # created empty dataframe for saving encoded values
for column in data.columns:
    dfe[column] = le.fit_transform(data[column])
dfe

enter image description here

# saving features
X = dfe[['code','sp']]

# saving target
y = dfe['res']

# splitting into training & test data
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=13)

X_train

enter image description here

# reversal of encoding
dfr_train = X_train.copy()
for column in X.columns:
    le.fit(data[column])   # you fit the column before it was encoded here

# now that python has the above encoding in its memory, we can ask it to reverse such 
# encoding in the corresponding column having encoded values of the split dataset

    dfr_train[column] = le.inverse_transform(X_train[column])
dfr_train

enter image description here

You can do the same for test data.

# reversal of encoding of data
dfr_test = X_test.copy()
for column in X.columns:
    le.fit(data[column])
    dfr_test[column] = le.inverse_transform(X_test[column])
dfr_test

Here is the full training data (features + variables) for export:

# reverse encoding of target variable y
le.fit(data['res'])
dfr_train['res'] = le.inverse_transform(y_train)
dfr_train     # unencoded training data, ready for export

enter image description here