🏃‍♀️ Customer Segmentation for Sportify Using Clustering Techniques

This project was developed for the Data Science and Machine Learning (DSML) 2023/2024 course. The goal was to apply unsupervised machine learning techniques to analyze and segment Sportify’s customer base using three datasets: Digital Contact, Product Purchases, and Demographics.

---

📁 Datasets Used

• Digital_Contact.csv: Engagement data (app, email, and social media interactions)
• Products.xlsx: Purchase history across product categories
• Demographic.txt: Customer attributes like name, birth year, city, education level

---

📌 Project Goals

• Clean and preprocess the datasets to prepare for analysis
• Apply clustering algorithms (K-Means, DBSCAN) to group customers
• Use dimensionality reduction (PCA, t-SNE) for visualization and improved clustering
• Describe clusters to identify customer profiles and recommend business actions

---

🧪 Methodology

🔍 Exploration

• Statistical and visual analysis of each dataset
• Key variables identified for clustering
• Creation of new variables (e.g., Days_Since_Last_Purchase, age, gender)

🧼 Preprocessing

• Normalization with MinMaxScaler
• Outlier detection and selective removal
• Missing value handling with KNNImputer and zero-fill

🤖 Modeling

• Clustering on Digital Contact and Product datasets:
  • K-Means (with and without PCA)
  • DBSCAN
  • t-SNE followed by K-Means or DBSCAN
• Cluster evaluation using:
  • Elbow method
  • Silhouette score
  • KL divergence (for t-SNE)

---

🔍 Cluster Profiles

💬 Digital Contact Segments

1. Curious Viewers – High email & SM clicks, low interactivity
2. Influencers – High SM likes, comments, shares
3. App Users – High app clicks, low SM engagement

🛍️ Product Purchase Segments

1. Low Spenders
2. Sport Lovers
3. Outdoor Enthusiasts
4. Occasional Customers

🧬 Combined Profiles (with Demographics)

• Influencers & Sport Lovers – Young, active, high spenders
• App Users & Occasional Buyers – Infrequent spenders
• Curious Viewers & Low Spenders – Older customers with minimal engagement
• … and more combinations based on cluster intersections

---

🎯 Business Recommendations

• Target Influencers via social media campaigns and loyalty rewards
• Upsell Outdoor Enthusiasts with bundled hiking gear
• Re-engage Curious Viewers via personalized email offers
• Incentivize App Users to become frequent buyers with app-only deals

---

📚 Tools & Libraries

• Python (Jupyter Notebook)
• pandas, numpy, scikit-learn, matplotlib, seaborn
• Algorithms: K-Means, DBSCAN, PCA, t-SNE
• Evaluation: Elbow method, Silhouette score, KL divergence

---

👥 Group Members

Group 18 – DSML 2023/2024

• Baran Can Çelik (20232067)
• Carlos Lourenço (20232020)
• Kida Aly (20231491)
• Priyá Dessai (20232053)
• Raysa Rocha (20232051)

---

📄 Files

📁 Sportify-Clustering-Group18/
├── DSML202324_Cluster_Group18_Report.pdf
├── DSML202324_Cluster_Group18_Report.docx
├── DSML202324_Cluster_Group18_Notebook.ipynb
├── README.md