Credit Card Customer Segmentation & Classification

📌 Overview

This project focuses on analyzing Credit Card Dataset for Clustering.
The workflow is divided into two main parts:

1. Unsupervised Learning (K-Means Clustering)
   • Used to segment customers based on their credit card usage patterns.
   • Achieved 0.32 Silhouette Score for Clustering 🎯
2. Supervised Learning (Random Forest Classification)
   • Built a classification model on the clustered dataset to predict cluster labels.
   • Achieved 98.88% accuracy 🎯.

📂 Dataset

• Source: Credit Card Dataset for Clustering (Kaggle)
• Description: Contains anonymized credit card usage data for Customers.

• Features:
  • TENURE
  • BALANCE_RANGE
  • PURCHASES_RANGE
  • ONEOFF_PURCHASES_RANGE
  • INSTALLMENTS_PURCHASES_RANGE
  • CASH_ADVANCE_RANGE
  • CREDIT_LIMIT_RANGE
  • PAYMENTS_RANGE
  • MINIMUM_PAYMENTS_RANGE
  • BALANCE_FREQUENCY_RANGE
  • PURCHASES_FREQUENCY_RANGE
  • ONEOFF_PURCHASES_FREQUENCY_RANGE
  • PURCHASES_INSTALLMENTS_FREQUENCY_RANGE
  • CASH_ADVANCE_FREQUENCY_RANGE
  • PRC_FULL_PAYMENT_RANGE
  • PURCHASES_TRX_RANGE
  • CASH_ADVANCE_TRX_RANGE

🛠️ Methodology

1. Data Preprocessing

• Handled missing values.
• Standardized numerical features.
• Prepared dataset for clustering.

2. Clustering with K-Means

• Chose optimal K using the Elbow method.
• Segmented customers into clusters.
• Visualized clusters for interpretation.

3. Classification with Random Forest

• Used K-Means cluster assignments as pseudo-labels.
• Trained a Random Forest Classifier to predict cluster membership.
• Evaluated performance with accuracy and classification report.

📊 Results

• Silhouette Score for Clustering: 0.32 ✅
• Random Forest Accuracy: 98.88% ✅
• Classification Report:
  • Precision, Recall, and F1-score all ≈ 0.99.
• Strong evidence that clusters are well-separated and predictable.

📈 Visualizations

_{Determining the optimal number of clusters using the WCSS (Elbow) technique.}

---

_{Visualization of the dataset after applying K-Means clustering.}

---

Developed as an Independent Machine Learning Project