In this project I trained machine learning and deep learning models in order to classify tweets as Neutral, Pro-vax or Anti-vax, experimenting with different configurations and metrics.
The models and their best accuracy scores:
- Logistic Regression: 0.7230
- Feedforward Neural Network: 0.6520
- LSTM: 0.7050
- GRU: 0.7063
- Bert pretrained model: 0.7751
Concepts examined:
- Data processing
- Feature extraction
- Data visualization
- Skip Connections
- Attention layer
- Activation functions
- Loss functions