This is a Python-based desktop application built during Kaufland’s Data Academy in collaboration with the University of National and World Economy (UNWE). The goal of the project was to solve a real business case:
“How can we help Kaufland choose higher-quality red wine for their stores?”
The application leverages machine learning models to predict the quality of red wine samples based on physicochemical characteristics. It includes features for data visualization, prediction, and model analysis — all wrapped in an intuitive PyQt5 interface.
- Python 3
- Pandas & NumPy
- Scikit-learn (RandomForestRegressor, DecisionTreeClassifier)
- Matplotlib & Seaborn
- PyQt5 (Desktop GUI)
- 📂 Import red wine dataset (CSV)
- 📊 Data inspection and visualization
- 📈 Train & analyze ML models (classification + regression)
- 🧠 Predict wine quality using AI
- ✅ Identify top-quality wines (true positives)
- 🌳 Visualize decision tree logic
- 📃 Export results and explanations
- 1,599 samples of red wine
- 11 physicochemical attributes
- Quality ratings (0–10) from professional sommeliers
Source: UCI Wine Quality Dataset
-
Random Forest Regressor
Trained to predict numeric wine quality scores and identify the most important features. -
Decision Tree Classifier
Used to classify wines as "Good" or "Not Good" based on quality threshold and to visualize decision paths.