Use the beer review data to do ONE of the following:
Normalize the 'rating' field to be between 0 and 1.
Create a logistic model that predicts a normalized review score between 0 and 1.
Note: this is NOT a classifier, it does not make sense to report a confusion matrix or % correct here.
You can report your accuracy by MSE, or something similar. Start off by normalizing reviews, creating text features, and reducing matrix size.
OR
Pick a style (with more than 500 reviews) or a group of styles.
Create a logistic classifier that predicts if a review is pertaining to that style.
Note: This IS a classifier and it would make sense to report a confusion matrix or % correct here.
Start off by normalizing reviews, creating text features, and reducing matrix size.
Beer review data is available on the github site OR here:
https://www.dropbox.com/s/3jlokbq7tjnbyr2/beer_reviews.csv?dl=0 (Links to an external site.)