Wine Quality Mining

• Data Warehousing and Data Mining Project

Project Objective

The project focuses on:

• understanding the structure and behavior of the wine dataset
• performing common preprocessing steps
• predicting wine quality categories using classification algorithms
• discovering hidden wine groups using clustering techniques
• comparing model behavior and interpreting results

Dataset

• Dataset used: winequality-red.csv
• Location: data/raw/winequality-red.csv
• Total columns: 12
• Target column: quality
• Input features: physicochemical properties such as acidity, chlorides, sulphates, alcohol, pH, density, and sulfur dioxide measures

Project Workflow

The project is divided into four major stages:

1. Exploratory Data Analysis

   • inspect dataset shape, data types, and summary statistics
   • analyze missing values, duplicates, and outliers
   • study quality distribution and feature relationships
   • generate EDA plots

2. Preprocessing

   • check missing values
   • remove duplicate rows
   • handle outliers
   • apply feature scaling
   • save processed datasets

3. Classification

   • treat quality as a multiclass target
   • split data into training and testing sets
   • train and compare classifiers
   • evaluate using accuracy, classification report, and confusion matrix
   • save the best trained model

4. Clustering

   • remove quality during clustering
   • standardize features
   • apply K-Means clustering
   • choose the best number of clusters using inertia and silhouette score
   • visualize clusters using PCA

Notebooks

• 01_EDA.ipynb Performs exploratory data analysis and generates plots for distribution, outliers, feature relationships, and correlation.

• 02_Preprocessing.ipynb Performs common preprocessing steps and saves cleaned and scaled datasets.

• 03_Classification.ipynb Performs multiclass classification, compares models, saves the best model, reloads it, and predicts a sample wine.

• 04_Clustering.ipynb Performs K-Means clustering, cluster selection analysis, PCA visualization, and clustered data export.

Models and Techniques Used

Classification

• Logistic Regression
• K-Nearest Neighbors
• Random Forest

Clustering

• K-Means Clustering
• PCA for 2D visualization

Evaluation

• Accuracy
• Classification Report
• Confusion Matrix
• Inertia
• Silhouette Score

Important Outputs

Processed Data

• winequality_cleaned.csv
• winequality_scaled.csv
• winequality_clustered.csv

Saved Model

• best_classification_model.pkl

Plots

• EDA plots:

  • quality_distribution.png
  • feature_distributions.png
  • feature_boxplots.png
  • feature_vs_quality_boxplots.png
  • correlation_heatmap.png

• Classification plots:

  • confusion_matrix_best_model.png

• Clustering plots:

  • cluster_selection_metrics.png
  • kmeans_pca_clusters.png

Project Structure

Wine-Quality-Mining/
├─ data/
│  ├─ raw/
│  └─ processed/
├─ models/
├─ notebooks/
│  ├─ 01_EDA.ipynb
│  ├─ 02_Preprocessing.ipynb
│  ├─ 03_Classification.ipynb
│  └─ 04_Clustering.ipynb
├─ plots/
│  ├─ Eda/
│  ├─ classification/
│  └─ clustering/
├─ requirements.txt
└─ Readme.md

How to Run

1. Open the project in the same virtual environment used for this repository.
2. Run the notebooks in order:
   • 01_EDA.ipynb
   • 02_Preprocessing.ipynb
   • 03_Classification.ipynb
   • 04_Clustering.ipynb
3. Check the generated outputs in:
   • data/processed
   • plots
   • models