YouTube Sentiment Analyzer (NLP Final Project)

This project analyzes YouTube comments and predicts sentiment as Positive, Negative, or Neutral using a TF-IDF vectorizer and a machine learning classifier.

Repository Requirements Checklist

This repository includes all required items:

• Python code files: app.py, train_model.py
• Dataset files: All_Comments_Final.csv, Aggregated_Metrics_By_Video.csv, Aggregated_Metrics_By_Country_And_Subscriber_Status.csv, Video_Performance_Over_Time.csv
• Trained model files: sentiment_model.pkl, tfidf_vectorizer.pkl
• GUI application code: app.py (Streamlit web app)
• Colab notebook: Youtube_Channel_Analysis_NLP_Project_C_final.ipynb

Project Structure

• app.py: Streamlit GUI for live sentiment prediction
• train_model.py: Model training script and pickle export
• sentiment_model.pkl: Trained sentiment model
• tfidf_vectorizer.pkl: Trained TF-IDF vectorizer
• All_Comments_Final.csv: Main comment dataset
• Youtube_Channel_Analysis_NLP_Project_C_final.ipynb: Analysis notebook (Colab compatible)
• Youtube_Channel_Analysis_NLP_Project_C_final.pdf: Project report/exported notebook

Setup

1. Create and activate a virtual environment.
2. Install dependencies:

pip install streamlit scikit-learn pandas textblob

Run the Web Application

python -m streamlit run app.py

Then open the local URL shown in the terminal (usually http://localhost:8501).

Retrain the Model (Optional)

If model files are missing or corrupted, retrain with:

python train_model.py

This regenerates:

• sentiment_model.pkl
• tfidf_vectorizer.pkl

Colab Usage

You can open Youtube_Channel_Analysis_NLP_Project_C_final.ipynb directly in Google Colab by uploading it to Colab or selecting it from your GitHub repository once pushed.

Notes

• The Streamlit app loads model files from the repository root.
• If pickle loading errors appear, run train_model.py to regenerate model artifacts.