Leverage Bayes' Theorem to create a supervised machine learning algorithm, thanks to sklearn.

naive_bayes_classifier_ex1.py

# REQUIREMENTS
# - A tagged dataset is necessary to calculate the probabilities used in Bayes' Theorem.
# - In order to apply Bayes' Theorem, we assume that these features are independent.
# - Using Bayes' Theorem, we can find P(class|data point) for every possible class. The class with the highest probability will be the algorithm’s prediction.

# MORE Improvements for the Natural Language Processing.
# - Remove punctuation from the training set (great! into great)
# - Lowercase every word in the training set (Great into great)
# - Use a bigram or trigram model, makes the assumption of independence more reasonable ("this is" or "is great" instead of single words)

from reviews import baby_counter, baby_training, instant_video_counter, instant_video_training, video_game_counter, video_game_training
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB

#The review to check
review = "Having sex is good"

#Tranforming the review for different kind of classifications
baby_review_counts = baby_counter.transform([review])
instant_video_review_counts = instant_video_counter.transform([review])
video_game_review_counts = video_game_counter.transform([review])

#initialize the classifiers
baby_classifier = MultinomialNB()
instant_video_classifier = MultinomialNB()
video_game_classifier = MultinomialNB()

#We have datapoints but we don't have labels which should be negative for the first 1000 and positive for the second 1000
baby_labels = [0] * 1000 + [1] * 1000
instant_video_labels = [0] * 1000 + [1] * 1000
video_game_labels = [0] * 1000 + [1] * 1000

#Fit the classifiers with (data points, data labels)
baby_classifier.fit(baby_training, baby_labels)
instant_video_classifier.fit(instant_video_training, instant_video_labels)
video_game_classifier.fit(video_game_training, video_game_labels)

#Predict the probability (with .predict_proba method) that the review is bad or good
print("Baby training set: " +str(baby_classifier.predict_proba(baby_review_counts)))
print("Amazon Instant Video training set: " + str(instant_video_classifier.predict_proba(instant_video_review_counts)))
print("Video Games training set: " + str(video_game_classifier.predict_proba(video_game_review_counts)))