kevinrpan / yummly

Machine Learning Project

Yummly

Machine Learning Project

Exploring Recipe Data

• Naive Bayes
• SVM
• Decision Trees

###lr.py Word lemmatization version 1 - better for logistic regression

cleaned up words and tfidf vectorized ingredients from the corpus

benchmark 50-50 split:

• NB .715
• LR .777
• SVC .778

###nb2.py Word lemmatization version 2 - better for NB simple lowercase lemmatization

benchmark 50-50 split:

• NB .739
• LR .774
• SVC .774

###nb.py no lemmatization.

Includes code to do kfold testing. Do not run kfold with gridsearch (slow)

benchmark 50-50 split:

• NB .731
• LR .729
• SVC .772