🌍 Global Development Categorization Dashboard

An end-to-end unsupervised machine learning project that clusters countries by socio-economic indicators and visualizes global development patterns through an interactive dashboard.

🔗 Live Demo: global-development-jpa3hswfotyh3mxnkocgfg.streamlit.app

---

📌 Project Overview

This project analyzes world development data across 22 socio-economic indicators — including GDP, health expenditure, CO2 emissions, internet usage, and population metrics — to automatically group countries into development categories using K-Means clustering.

The goal is to answer: Can we objectively classify countries by development level using only data?

---

🖼️ Dashboard Preview

World Map	Country Analysis
Choropleth map colored by development level	Radar chart comparing a country vs its cluster average

---

⚙️ Tech Stack

Layer	Tools
Language	Python 3
ML	Scikit-Learn (KMeans, SimpleImputer, StandardScaler)
Data	Pandas, NumPy
Visualization	Plotly Express
App Framework	Streamlit
Deployment	Streamlit Cloud

---

📂 Project Structure

Global-Development/
│
├── app.py                          # Streamlit dashboard
├── train_model.py                  # Preprocessing pipeline & artifact generation
├── World_development_mesurement.xlsx  # Raw dataset
├── requirements.txt                # Dependencies
│
└── model/
    ├── imputer.pkl                 # Fitted SimpleImputer
    ├── scaler.pkl                  # Fitted StandardScaler
    └── features.pkl                # Feature column names

---

🧠 ML Pipeline

Raw Data
   │
   ▼
Special Character Removal  ──  ($, %, , in GDP / Tax Rate / Tourism cols)
   │
   ▼
Drop Irrelevant Columns  ──  (Ease of Business >50% missing, Number of Records)
   │
   ▼
Median Imputation  ──  (SimpleImputer — handles remaining nulls)
   │
   ▼
Log Transformation  ──  (np.log1p on 11 right-skewed columns)
   │
   ▼
Standard Scaling  ──  (StandardScaler — zero mean, unit variance)
   │
   ▼
K-Means Clustering  ──  (k=2 to 6, user-controlled via sidebar)
   │
   ▼
GDP-Based Label Assignment  ──  (Under-Developed → Developing → Developed ...)

---

📊 Features

• Interactive World Map — choropleth colored by development level
• Adjustable Clusters — sidebar slider lets you change k from 2 to 6 live
• Country Drilldown — zoom into any individual country on the map
• Radar Chart — compare a selected country's scaled indicators vs its cluster average
• Distribution Plot — box plot of any indicator split by development level
• Single Country Analysis — indicator-by-indicator comparison vs global average with status labels (Good / Average / Needs Improvement)
• CSV Download — download a full development report for any country

---

📈 Key Results (k=3)

Cluster	Label	Characteristics
0	Under-Developed	Low GDP, high birth/infant mortality rate, low internet usage
1	Developing	Mid-range GDP, growing mobile/internet penetration
2	Developed	High GDP, high health expenditure, low mortality rates

Silhouette Score at k=3: ~0.28 (meaningful separation given 22-dimensional data)

---

🚀 Run Locally

# 1. Clone the repo
git clone https://github.com/JAGGU-528/Global-Development.git
cd Global-Development

# 2. Install dependencies
pip install -r requirements.txt

# 3. Generate model artifacts (only needed once)
python train_model.py

# 4. Launch the dashboard
streamlit run app.py

---

📋 Dataset

World Development Measurement — contains 2,700+ rows across 195+ countries with 24 original features including:

• Birth Rate, Infant Mortality Rate, Life Expectancy (Male/Female)
• GDP, Health Exp/Capita, Health Exp % GDP
• CO2 Emissions, Energy Usage
• Internet Usage, Mobile Phone Usage
• Tourism Inbound/Outbound
• Population (Total, Urban, Age groups)
• Busi
• ness Tax Rate, Days/Hours to Start Business, Lending Interest

---

💡 What I Learned

• How to build a complete ML pipeline from raw messy data to a deployed web app
• Why preprocessing order matters — impute before log transform, not after
• How pickle version pinning is critical for reproducible deployments
• How to use @st.cache_data and @st.cache_resource correctly in Streamlit
• Practical difference between K-Means, DBSCAN, Hierarchical, and GMM clustering

---

👤 Author

JAGDISH BIRADAR — Electronics & Communication Engineering Graduate
Targeting Data Science / ML Engineer roles.

---

Built as part of a structured Data Science learning roadmap. All preprocessing, model training, deployment debugging, and dashboard development done independently.