An AI-powered machine learning system that automatically detects toxic comments on YouTube using Natural Language Processing (NLP) and multiple classification algorithms.

• The dataset used is Jigsaw Toxic Comment Classification (Kaggle)
• The dataset consists of 21,825+ comments categorized under 6 Labels : Toxic, Severe Toxic, Obscene, Threat, Insult, Identity Hate

• @rishi shandilya

• Python 3.x
• Pandas & NumPy - Data manipulation
• Matplotlib & Seaborn - Data visualization
• NLTK - Text preprocessing
• Scikit-learn - Machine Learning
• Google Colaboratory - Development environment

1. Data Loading & Exploration

   • Load dataset and analyze distribution
   • Visualize toxicity patterns

2. Data Preprocessing

   • Text cleaning (remove URLs, special characters)
   • Stopword removal
   • Tokenization

3. Feature Engineering

   • TF-IDF vectorization
   • Convert text to numerical features

4. Model Training

   • Logistic Regression
   • Random Forest Classifier
   • Naive Bayes

5. Model Evaluation

   • Accuracy comparison
   • Confusion matrices
   • Performance metrics

• Best Model: Logistic Regression
• Accuracy: 95%+
• Training Time: <10 seconds