White Wine Quality Classification

Project Overview

This project explores classification techniques in data mining to model the taste preferences for white wine based on easily accessible analytical data during any wine's certification phase. The best-performing model in our analysis was the K-Nearest Neighbors (K-NN) trained on a balanced set, showing a commendable balance between sensitivity and specificity.

Data

The dataset titled "White Wine Quality" consists of 4989 observations of Vinho Verde wine variant physical-chemical characteristics from 2004 to 2007, including median taste preference scores from blind tasting sessions.

Methodology

• Preliminary data analysis for missing values, outliers, and variable distribution.
• Classification models like K-Nearest Neighbors (K-NN), Logistic Regression, and Discriminant Analysis.
• Downsampling technique to address class imbalance.
• Performance evaluation using metrics like Accuracy, AUC, and Balanced Accuracy.

Results

Our models achieved varying degrees of success, with the balanced K-NN model performing best in terms of balanced accuracy, indicating a successful classification of white wine quality.

How to Run

1. Ensure R is installed on your system.
2. Clone this repository.
3. Run the code.R script to perform the analysis.

Authors

• Julius Maliwat
• Filippo Bianchini
• Giacomo Rabuzzi
• Andrea Robbiani