Severity-of-Toxic-Comments-MLOps

Methodology

• Data Analysis
• Data preprocessing
• Text Normalisation
• Lemmatization
• Stop-words Removal
• Tokenization
• Embedding words into vectors using FastText
• Trained Model using LSTM, LSTM-CNN
• Model Evaluation
• Achieved the best accuracy using LSTM

Workflow - deployment

• Saved the tokenizer and the DL model.
• Developed backend and frontend using Python and Streamlit library, deployed the web app on localhost.
• Containerized the software using Docker and tested it on localhost.
• Hosted EC2 instance on AWS and installed Docker. I connected the EC2 to GitHub using GitHub actions for the CI/CD pipeline.
• Used Elastic Container Registry (ECR) to store the Docker image
• Acurracy : Train Data - 99.08% Test Data - 99.32%