3) Data preparation (CRISP-DM methodology)

Cleaning:

• Converted TotalCharges to numeric.
• Removed rows with missing values.
• Dropped the customerID column (irrelevant).

Encoding:

• Categorical features encoded using LabelEncoder.
• The target Churn was mapped to binary (0 = No, 1 = Yes).

Balancing:

• Used SMOTE to address class imbalance.
• Scaling:
• Applied StandardScaler to normalize numerical features.

---

⬅ Voltar para a Página Inicial