twitter_sentiment_analysis

Comparing various classifier models to find the most accurate model for Twitter Sentiment Analysis.

Dataset:

Dataset conatins train.csv and test.csv contain 29,530 and 16,130 tweets respectively provieded by Kaggle.

Requirements:

Some of the general library requirements for this project are:

1. numpy
2. nltk
3. pandas
4. scikit-learn
5. regex
6. tweepy

Libraries specific to some methods are:

1. keras with TensorFlow backend for RNN(bidirectional LSTM)
2. scikit-learn
3. xgboost for XGBoost
4. nltk for vader sentiment analysing.

Models used

1. Naive Bayes
2. K-neighbour
3. Random Forest
4. Logistic
5. XG Boost
6. SVM
7. RNN(bidirectional LSTM)
8. Vader Pre-trained model
9. BERT
10. ULMFiT