Access the web app : https://winequalityimbalancedlearnign.herokuapp.com (Currently down)
Prediction of wine quality based on physical/chemical features, with Imbalanced Learning techniques.
The frequencies of each quality class are shown in the figure below:
As we can see, the lowest and highest classes have very small frequencies, compared to the medium qualities. This could indicate an Imbalanced Learning problem. Therefore, some special technicques were performed:
- data pre-processing (scaling)
- oversampling (SVMSMOTE)
- weighted class in Random Forest
To run the model, execute in the terminal of the root directory:
$ streamlit run app.py
A webpage will be open in a localhost, with the web app.
To stop: CRTL+C.