Wine & Quality Type Predictor

Abstract:

Roughly 60% of Americans claim to enjoy an occasional glass of wine, whether it’sat a special dinner or at home. Two of the most popular wines are red wine andwhite wine, which more than 60% of wine consumers drink. Both red and white wines, while strikingly different in appearance, have similar distributions of pH, sulfuric oxides, chlorides, and other elements. The goal of our project is to apply machine learning algorithms to correctly classify a red or white wine sample and to predict wine quality. A dataset provided by UCI was used for our statistical analyses. Preliminary steps included preprocessing the data into the desired form, and then applying, PCA (Principal Component Analysis), an unsupervised learning algorithm to find trends in our unstructured data. Afterwards, we implemented three supervised learning algorithms (Polynomial Regression, Logistic Regression, and Support Vector Machines) to test the accuracy of the predicted results. Another goal of our project was to illustrate the effect of regularization and feature transformation on our data. Lasso (L1) and Ridge (L2) regularization were applied to the linear and logistic regression models. Three different kernels were used and compared within our SVM model.

Paper:

https://www.overleaf.com/1871767983rcmkbjwnhbnp

PowerPoint:

https://nyu0-my.sharepoint.com/:p:/g/personal/grc284_nyu_edu/Ef8IhTjOusNJjM8gqVGp4bwBEetQkCiWu1DMNVReORzYeg?e=ROKYGn

Dataset (from UCI Machine Learning Repository)

HTTPS://ARCHIVE.ICS.UCI.EDU/ML/DATASETS/WINE+QUALITY