- As final project from course provided by Kagglemaster Mario Filho, the target of this project is to implement a bunch of shortcuts and best practices learned in his course.
- From data extraction to deploying model, the project uses an ensemble model (LightGBM and RandomForest) to predict which ytube video would be interesting based on three keywords ["machine+learning", "data+science", "kaggle"].
Project link: https://mltubeapp.herokuapp.com/
- Data Extraction
- EDA
- Modeling
- Simple Front Page
- Heroku Deploy
- API Prediction
- Data Extractor broken! - Fix Data Extractor
- Implement some Database
- Implement Cloud of words chart based on TFIDVectorizer
- gunicorn==20.0.4
- Flask==1.1.2
- requests==2.25.0
- beautifulsoup4==4.9.3
- pandas==1.1.4
- joblib==0.17.0
- numpy==1.19.4
- scipy==1.5.4
- scikit-learn==0.23.2
- lightgbm==2.3.0
Running local with Flask:
- Git clone repository
- Create venv
- Activate venv
- Install requirements
- Start Flask server
git clone https://github.com/fduque/youtube_recommender_ML_app.git ytube
cd ytube
python -m venv .ytube
source .ytube/bin/activate
pip install -r requirements.txt
flask run
Running local with Docker:
- Turn on Docker Desktop
- Execute build docker
- Run docker
docker build . -t appinstance1
docker run -e PORT=80 -p 80:80 appinstance1