MLOps Project: Hate Speech Detection API

This project demonstrates a complete MLOps workflow for building, improving, and containerizing a machine learning model. The goal is to detect hate speech in text and serve the model via a robust FastAPI.

Key Achievements

• Data Analysis: Performed exploratory data analysis on an imbalanced dataset, identifying the need for resampling.
• Model Improvement: Successfully addressed class imbalance using SMOTE to significantly improve hate speech detection recall from 17% to 54%.
• API Development: Created a RESTful API with FastAPI to serve the model predictions.
• Containerization: Packaged the entire application in a Docker container for portability and deployment.

Technologies

• Python
• Scikit-learn
• FastAPI
• Docker

---

How to Run the Project 🏃‍♂️

You can run this project locally with Docker.

1. Clone the repository:

   git clone <your_github_repository_url>
   cd <your_project_name>

2. Build the Docker image:

   docker build -t hate-speech-api .

3. Run the Docker container:

   docker run -d -p 8000:8000 --name hate_speech_detector hate-speech-api

4. Access the API: Your API is now running at http://localhost:8000. You can test it by going to the interactive docs: http://localhost:8000/docs.

---

Model Performance 📊

The final, optimized model provides a strong balance of overall accuracy and high recall on the critical hate speech class.

• Final Accuracy: 85.27%
• Hate Speech Recall (Class 0): 54%
• Hate Speech Precision (Class 0): 29%
• Hate Speech F1-Score (Class 0): 0.38