Identifying the most positive and negative words
In this exercise we'll try to interpret the coefficients of a logistic regression fit on the movie review sentiment dataset. The model object is already instantiated and fit for you in the variable lr.
In addition, the words corresponding to the different features are loaded into the variable vocab. For example, since vocab[100] is "think", that means feature 100 corresponds to the number of times the word "think" appeared in that movie review.
This exercise is part of the course
Linear Classifiers in Python
Exercise instructions
- Find the words corresponding to the 5 largest coefficients.
- Find the words corresponding to the 5 smallest coefficients.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Get the indices of the sorted cofficients
inds_ascending = np.argsort(lr.coef_.flatten())
inds_descending = inds_ascending[::-1]
# Print the most positive words
print("Most positive words: ", end="")
for i in range(5):
print(____, end=", ")
print("\n")
# Print most negative words
print("Most negative words: ", end="")
for i in range(5):
print(____, end=", ")
print("\n")