Big-Data-Analytics

Unsupervised Clustering & Country Segmentation using RapidMiner

Unsupervised Clustering of Countries Using Economic & Business Indicators

Project Overview

This project applies unsupervised machine learning (clustering) to group countries based on 25 indicators related to transportation capacity, business environment, financial development, and socioeconomic status.

The objective is to identify homogeneous groups of countries and determine the optimal number of clusters using statistical performance metrics.

---

Dataset Description

• 25 country-level attributes
• Covers:
  • Transportation & logistics capacity
  • Business environment & entrepreneurship
  • Financial and economic indicators
  • Socioeconomic development metrics
• Countries include developed, developing, and least developed economies

---

Methodology

• Data standardization
• K-Means clustering
• Evaluation using:
  • Silhouette Score
  • Davies–Bouldin Index (DBI)
• Models tested with:
  • K = 2 clusters
  • K = 3 clusters

---

Key Findings (Business Interpretation)

1. Natural Segmentation of Countries

The clustering exercise clearly grouped countries into:

• High-capacity / developed economies
• Lower-capacity / developing & least-developed economies

This confirms that the dataset captures meaningful global economic differences.

---

2. Transportation Capacity as a Major Driver

Indicators such as:

• Air freight movement
• Air passenger traffic
• Container port volume

were among the strongest contributors to cluster separation, highlighting the importance of logistics infrastructure in economic development.

---

3. Business Environment Matters

Variables such as:

• Cost & time to start a business
• Startup density
• Number of procedures

helped separate business-friendly economies from restrictive ones.

---

4. USA and China in Same Cluster (K = 2)

In the 2-cluster solution, USA and China appeared in the same high-capacity cluster due to:

• Large-scale economic output
• Strong transportation networks
• High business activity levels

This shows clustering captures capacity, not income level alone.

---

Clustering Performance Evaluation

K = 2 Clusters

• Silhouette Score: 0.639 (high separation)
• Davies–Bouldin Index: 0.717 (compact clusters)

➡️ Indicates strong, well-separated clustering.

---

K = 3 Clusters

• Silhouette Score: 0.300 (weak separation)
• Davies–Bouldin Index: 1.101 (overlapping clusters)

➡️ Clusters are less distinct and statistically weaker.

---

Final Conclusion

Although K = 3 produces intuitive labels (developed, developing, least-developed), K = 2 is statistically optimal for this dataset.

The performance indicators clearly support a two-cluster structure as the most meaningful and robust segmentation.

---

Tools & Skills Demonstrated

• Unsupervised Machine Learning (K-Means)
• Cluster evaluation metrics
• High-dimensional data analysis
• Business & economic interpretation
• Excel-based analytical modeling

---

Business Relevance

Such clustering can be used for:

• Market entry strategy
• Global investment prioritization
• Infrastructure planning
• Policy & development analysis