I have trained a Logistic Regression classifier to predict whether a review is positive or negative. Now, I want to append the predicted probabilities returned by the predict_proba-function to my Pandas data frame containing the reviews. I tried doing something like:
test_data['prediction'] = sentiment_model.predict_proba(test_matrix)
Obviously, that doesn't work, since predict_proba returns a 2D-numpy array. So, what is the most efficient way of doing this? I created test_matrix with SciKit-Learn's CountVectorizer:
vectorizer = CountVectorizer(token_pattern=r'\b\w+\b')
train_matrix = vectorizer.fit_transform(train_data['review_clean'].values.astype('U'))
test_matrix = vectorizer.transform(test_data['review_clean'].values.astype('U'))
Sample data looks like:
| Review | Prediction |
| ------------------------------------------ | ------------------ |
| "Toy was great! Our six-year old loved it!"| 0.986 |
Assign the predictions to a variable and then extract the columns from the variable to be assigned to the pandas dataframe cols. If x is the 2D numpy array with predictions,
x = sentiment_model.predict_proba(test_matrix)
then you can do,
test_data['prediction0'] = x[:,0]
test_data['prediction1'] = x[:,1]
import numpy as np
import pandas as pd
df = pd.DataFrame(
np.arange(10).reshape(5, 2), columns=['a', 'b'])
print('df:', df, sep='\n')
arr = np.arange(100, 104).reshape(2, 2)
print('array to append:', arr, sep='\n')
df = df.append(pd.DataFrame(arr, columns=df.columns), ignore_index=True)
print('df:', df, sep='\n')
output
df:
a b
0 0 1
1 2 3
2 4 5
3 6 7
4 8 9
array to append:
[[100 101]
[102 103]]
df:
a b
0 0 1
1 2 3
2 4 5
3 6 7
4 8 9
5 100 101
6 102 103