Methodology:
• Use NLTK stop words and Post of Speech to deduct meaningless words in every tweet.
• Utilized the regular expression to delete emotional figure in each tweet.
• Use discourse connectors list to remove redundant tokens.
• Implement the Voting Mechanism to build the sentiment prediction model with three classifiers: Multinominal Naive Bayes(MNB), Stochastic Gradient Descent(SGD), and Bernoulli Naive Bayes(BNB).
Link: https://drive.google.com/open?id=1dmQ8YZHnf98cpMEiV9ogW2xDAa7J0FkY