🛑 Toxic Comment Classification

🧩 Problem

Online platforms receive thousands of comments — some of them harmful.
This project detects toxic behavior in text using machine learning.

---

🏷️ Labels

Each comment can belong to one or more categories:

• toxic
• severe_toxic
• obscene
• threat
• insult
• identity_hate

👉 This is a multi-label classification problem

---

🧠 Approach

• Convert text into numerical features
• Train models to predict multiple labels
• Use One-vs-Rest strategy for handling multi-label outputs

---

🤖 Models Used

• Logistic Regression
• Multinomial Naive Bayes
• Wrapped with OneVsRestClassifier

---

📊 Evaluation

• Accuracy Score
• ROC-AUC Score
• Classification Report
• ROC Curve

---

🛠️ Tools

• Python
• Scikit-learn
• NLP preprocessing (tokenization, cleaning)

---

▶️ Run

git clone https://github.com/your-username/toxic-comment-classification.git
cd toxic-comment-classification
pip install -r requirements.txt
python app.py