In the era of big data, generic marketing is obsolete. This project leverages Unsupervised Machine Learning to segment 2,240 customers into distinct personality groups.
By analyzing behavior (spending, income, family structure), we identified 4 key customer segments. This allows for hyper-personalized marketing strategies that can reduce acquisition costs and increase Customer Lifetime Value (CLV).
Note: This project complements my work in Industrial Quality Control (IQC), demonstrating my ability to apply statistical methods across both Manufacturing and Business Intelligence domains. data source:https://www.kaggle.com/datasets/imakash3011/customer-personality-analysis?select=marketing_campaign.csv
The workflow transforms raw data into actionable business insights:
- Advanced Feature Engineering:
- Derived
Customer_Tenure,Family_Size, andTotal_Spentto enrich the dataset.
- Derived
- Outlier Detection (Isolation Forest):
- Utilized
IsolationForestto remove statistical anomalies (top 1% noise), ensuring robust clustering.
- Utilized
- Dimensionality Reduction (PCA):
- Reduced 7 dimensions to 3 Principal Components using PCA, retaining variance while simplifying visualization.
- Clustering (K-Means):
- Segmented customers into 4 clusters.
- Validated using Silhouette Score.
Visualizing the separation of customer groups in PCA space.
A clear positive correlation, but with distinct behavioral boundaries.
Based on the cluster analysis, we recommend the following strategies:
| Cluster | Profile Description | Recommended Strategy |
|---|---|---|
| 0 (Red) | The Elite Spenders High Income, High Spending, Low Children. |
VIP Treatment: Offer exclusive early access to products and loyalty tiers. |
| 1 (Green) | Budget Families Lower Income, High Family Size. |
Volume Discounts: Promote "Family Packs" and BOGO (Buy One Get One) deals. |
| 2 (Blue) | New/Risk Customers Variable Income, Low Frequency. |
Retention Campaigns: Send personalized emails to re-engage before they churn. |
| 3 (Yellow) | Potential Upscalers Middle Income, Moderate Spending. |
Cross-Selling: Recommend higher-margin products based on purchase history. |
- Clone the repository.
- Install dependencies:
pip install -r requirements.txt - Run the modeling script:
python src/modeling.py
CHONG JIA YOU Bachelor of Science in Industrial Statistics, UTHM
This project was inspired by the work of Karnika Kapoor on Kaggle. Original Notebook: Customer Segmentation: Clustering 🛍️🛒🛒

