Wine Quality – Machine Learning Analysis

This project was developed during my Erasmus at the University of Latvia (Fall semester) as part of a Machine Learning module. It presents an end-to-end machine learning workflow applied to the Wine Quality dataset, covering both classification and regression approaches:

• Classification models to predict discrete quality levels
• Regression models to predict continuous quality scores
• Hyperparameter tuning and performance comparison

The objective was not only to train and optimise models, but also to critically analyse model behaviour, compare problem formulations (classification vs regression), and evaluate the impact of dataset structure on model performance and interpretability.

Overview

This project presents an end-to-end machine learning analysis of the Wine Quality dataset (UCI ML Repository). The objective is to explore both classification and regression approaches for predicting wine quality.

The project includes:

• Exploratory Data Analysis (EDA)
• Classification models (baseline + tuned)
• Regression models (baseline + tuned)
• Hyperparameter optimisation
• Model comparison and critical evaluation

Dataset

Wine Quality Dataset
Source: UCI Machine Learning Repository

Methods

• Logistic Regression
• Random Forest
• Support Vector Machines
• Gradient Boosting (if used)
• Hyperparameter tuning with GridSearchCV
• Evaluation metrics: Accuracy, F1-score, RMSE, MAE

Key Insights

• Comparison between classification and regression framing
• Impact of hyperparameter tuning
• Discussion of model suitability given dataset structure

Technologies in Google Colab used

• Python
• Scikit-learn
• Pandas
• NumPy
• Matplotlib / Seaborn