🍷 Wine Quality Prediction - Good or Not?

This project uses machine learning to predict whether a white wine is "good" or "not good" based on its physicochemical properties. The dataset is from the UCI Machine Learning Repository.

---

📁 Dataset Overview

The dataset includes 12 features:

• fixed acidity
• volatile acidity
• citric acid
• residual sugar
• chlorides
• free sulfur dioxide
• total sulfur dioxide
• density
• pH
• sulphates
• alcohol
• quality (score from 0 to 10)

---

🧪 Objective

The goal is to transform the quality column into a binary classification problem:

• ✅ Good Wine → quality >= 7 → Label: 1
• ❌ Not Good Wine → quality < 7 → Label: 0

Build a classification model to predict this label.

---

⚙️ Data Preprocessing

📌 1. Selected Features

Based on correlation analysis, we retained these five features for modeling:

features = ['alcohol', 'density', 'volatile acidity', 'chlorides', 'total sulfur dioxide']

🏷️ 2. Label Conversion

The binary target label was created as follows:

white['quality_label'] = (white['quality'] >= 7).astype(int)

⚠️ 3. Outlier Removal (Optional)

Used the IQR method to filter out outliers from numeric columns:

outlier_indices = set()
for col in white.select_dtypes(include='number').columns:
    Q1 = white[col].quantile(0.25)
    Q3 = white[col].quantile(0.75)
    IQR = Q3 - Q1
    lower = Q1 - 1.5 * IQR
    upper = Q3 + 1.5 * IQR
    outlier_indices.update(white[(white[col] < lower) | (white[col] > upper)].index)

white_clean = white.drop(index=outlier_indices)

---

🤖 Model Training

📊 Train/Test Split

from sklearn.model_selection import train_test_split

X = white[features]
y = white['quality_label']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

🧠 Logistic Regression Model

from sklearn.linear_model import LogisticRegression

model = LogisticRegression()
model.fit(X_train, y_train)

---

📈 Model Evaluation

from sklearn.metrics import accuracy_score, classification_report

train_pred = model.predict(X_train)
test_pred = model.predict(X_test)

print("Train Accuracy:", accuracy_score(y_train, train_pred))
print("Test Accuracy:", accuracy_score(y_test, test_pred))
print(classification_report(y_test, test_pred))

---

🔍 Predict from User Input

alcohol = float(input("Enter alcohol: "))
density = float(input("Enter density: "))
volatile_acidity = float(input("Enter volatile acidity: "))
chlorides = float(input("Enter chlorides: "))
total_sulfur = float(input("Enter total sulfur dioxide: "))

sample = [[alcohol, density, volatile_acidity, chlorides, total_sulfur]]
result = model.predict(sample)

if result[0] == 1:
    print("✅ The wine is GOOD")
else:
    print("❌ The wine is NOT GOOD")

---

📊 Visualizations

Created with Seaborn and Matplotlib:

• 📦 Boxplots to detect outliers
• 🔥 Correlation heatmap
• 🧮 Pairplots for multivariate relationships
• 📉 Scatter plots for visual trends

---

✅ Key Findings

• Alcohol shows the strongest positive correlation with wine quality.
• Logistic Regression performed well as a baseline model.
• Accuracy was reasonable on both train and test sets.

---

🚀 Future Improvements

• Try advanced models: Random Forest, XGBoost
• Apply hyperparameter tuning
• Add a Streamlit GUI for user-friendly predictions
• Deploy the model using Flask or FastAPI

---

📚 References

• UCI Wine Quality Dataset
• Scikit-learn Documentation
• Seaborn Documentation
• Matplotlib Documentation

---