About the code:

This is a sentiment analysis problem statement solved using the dataset from kaggle. https://www.kaggle.com/datatattle/covid-19-nlp-text-classification

The main aim was multiclass classification of tweets using NLP

Both Machine Learning and Deep Learning approaches were explored: ML models include:

• Multinomial Naive Bayes Classifier,
• Gradient Boosting classifier,
• Random Forest Classifier( relatively best training accuracy around 77%)
• As expected, they performed poorly on true test data(only 35% accuracy)

Deep Learning approach includes:

• A simple RNN model,(Accuracy: around 75%)
• A LSTM based model,(Accuracy around 82%)
• A Bidirectional LSTM model(Accuracy around 84% but suffers from overfitting)

How To Use

This folder contains 3 files:-

1. Sentiment Analysis.ipynb
2. Corona_NLP_train.csv
3. Corona_NLP_test.csv

----------About the module--------------

• Tools used: Jupyter notebook in Ananconda environment
• Dependencies: Python 3, Tensorflow version 2.5.0, Keras, nltk
• Libraries used: Numpy, Sklearn, Seaborn, Keras, Tensorflow, Matplotlib, gensim

=====Instructions to run the code======

1. IN JUPYTER NOTEBOOK:

• The folder contains the train and test data in form of .csv files ('Corona_NLP_train.csv' and 'Corona_NLP_test.csv')
• Ensure you download the whole folder and not change the relative path of .ipynb and .csv files.
• Run the code cells sequentially NOTE: Some models may take time to execute

2. In GOOGLE COLABORATORY

• open the .ipynb file
• Upload both the .csv files using the file upload option( mostly available at left hand side menu bar)
• Ensure upload is completed
• Execute cells sequentially NOTE: Some models may take time to execute