.wine

Inspiration

We were inspired by the connection between a subjective rating given by a human with the sample's physical and chemical properties.

What it does

Our project predicts the quality rating of red wine from a panel of human judges using different models with the given 11 physiochemical variables. Our leading question is to see if the ratings from the judges have any basis or connection with the physiochemical properties of the wine.

How we built it

We built this using the collaborative features of the SingleStore notebook. We also utilized sk-learn, pytorch, XGBoost for model building. We used matplotlib and seaborn for visualization.

Challenges we ran into

We had issues with determining the hyperparameter of our neural network given our attempt at including as much possible outputs for the model's predicted values.

Accomplishments that we're proud of

We are accuracy of our models created within this project given our relevant experience.

What we learned

We learned that predicting a variable from a small group of responses, a decision tree boosted by xg-boost was the best in terms of accuracy. We also learned that a complex method like a neural network was not ideal with this given dataset.