A data-driven exploration of color patterns, cultural identity, and post-colonial symbolism across all 54 African nations
- Research Question
- Key Findings
- Methodology
- Cluster Breakdown
- Visualizations
- Project Structure
- Quick Start
- Technologies
- Key Statistics
- Future Work
- License
- About
"Beyond shared colonial history, do flags with similar colors reflect deeper cultural, geographic, or ideological connections?"
Answer: YES. Color similarity correlates with geographic proximity, ideological movements, and design era far more strongly than colonial heritage alone. Post-independence Africa chose its own colors.
West African countries cluster together regardless of colonial power — French (50%), British (23%), and Portuguese (15%) colonies share the same color profile because they share a neighborhood and inter-connected independence movements.
The red-black-gold palette didn't spread uniformly. It came in three distinct waves:
| Wave | Decade | Count | Representative Countries |
|---|---|---|---|
| Pioneers | 1950s | 4 flags | Ghana, Guinea |
| Uptake | 1960s | 5 flags | Diverse choices during the independence rush |
| Resurgence | 1990s | 8 flags | Democratic transitions, renewed solidarity |
- 1960s: Simple horizontal or vertical tricolor stripes
- 1990s: Complex diagonals, Y-shapes, and radiating patterns (South Africa, Rwanda)
It matters — but geography, ideology, and timing override it. A country's neighbors and political alignment predict its flag colors better than its former colonizer.
The only country in its own solo cluster. Tunisia maintained its Ottoman-era crescent-and-star aesthetic while the rest of the continent moved toward Panafricanist or Socialist palettes. Its isolation validates the model — if clustering were random, outliers would not be this historically explainable.
- 54 African countries, manually verified
- 39 features per country: geography, colonial history, regional economic blocs, flag metadata, symbols, stripe pattern type, year of adoption vs. year of independence
- Adaptive K-Means clustering (1–6 colors per flag, elbow method)
- ~200 unique colors extracted total
- 1% minimum area threshold to filter micro-detail noise
- RGB color space with proportion weighting (dominant colors carry more weight)
- Hierarchical clustering with average linkage
- Bidirectional weighted Euclidean distance (color distance + proportion distance)
- Silhouette score optimization across k = 2–12 → 7 optimal clusters
- Silhouette score: 0.260 (modest — appropriate for culturally complex data)
- Cross-referenced with historical independence timelines
- Geographic pattern verification against African Union regions
- Temporal analysis: flag adoption year (not independence year — these differ for several countries)
| # | Name | Size | Dominant Colors | Notes |
|---|---|---|---|---|
| 1 | Blue-Green Simplicity (British Anglophone) | 3 |  Blue, Green, White |
Former British territories with simple bicolor or tricolor designs |
| 2 | Blue-Dominant Outliers | 2 |  Light Blue, White, Black |
Unique palettes dominated by blue — geographic and aesthetic outliers |
| 3 | Central-East Africa Pattern | 4 |  Green, Red, Black |
East/Central African flags with strong black presence |
| 4 | Panafricanist Movement & Complex Patterns | 26 |  Gold, Red, Green, Black |
Largest cluster — the Pan-African palette (Ethiopian colors) |
| 5 | Mixed Regional Patterns | 9 |  Orange, varied |
Diverse designs that don't fit a single ideological pattern |
| 6 | Socialist Liberation (Red-Black-Gold) | 9 |  Crimson, Black, Gold |
Flags shaped by Marxist liberation movements |
| 7 | North African Outlier (Tunisia) | 1 |  Red, White |
Solo Ottoman-influenced outlier |
All flags organized by color cluster. The visual coherence within each group validates the quantitative clustering.
The 5 most representative colors for each cluster, weighted by surface area across all member flags.
Regional patterns become immediately visible — West Africa's coherence, North Africa's divergence, and the Pan-African belt across the continent.
Three waves of Panafricanist adoption (1950s, 1960s, 1990s) overlaid on the full adoption timeline. The 1960s independence peak is annotated.
One-page visual summary: primary finding, cluster distribution, correlation strengths, and key insights. Portfolio and presentation ready.
A fully interactive HTML report with animated stats, cluster exploration modals, flag hover metadata, country search, and a scroll-animated SVG timeline.
Works offline — open directly in any modern browser, no server required.
african_flags_project/
├── notebooks/
│ ├── 01_data_collection.ipynb # Data gathering, enrichment, 39-feature dataset
│ ├── 02_color_extraction.ipynb # Adaptive K-Means color extraction (~200 colors)
│ ├── 03_color_similarity_clustering.ipynb # Hierarchical clustering, silhouette optimization
│ └── 04_visualizations.ipynb # Publication-quality plots + infographic
├── data/
│ ├── raw/
│ │ ├── african_countries_info.csv # Base country metadata
│ │ ├── african_countries_enhanced.csv # Full 39-feature dataset
│ │ ├── african_flags_with_colors.csv # Extracted color data
│ │ ├── african_flags_complete.csv # Combined dataset
│ │ ├── african_flags_final.csv # Cleaned final dataset
│ │ ├── african_flags_with_clusters.csv # Post-clustering assignments
│ │ ├── african_flags_clustered.csv # Primary analysis dataset
│ │ ├── cluster_summary.csv # Per-cluster statistics
│ │ └── cluster_analysis_summary.csv # Cluster interpretation notes
│ ├── flags_images/ # 54 PNG flag images (Country_Name.png)
│ └── visualizations/
│ ├── 01_cluster_flag_grid.png
│ ├── 02_cluster_color_palettes.png
│ ├── 03_geographic_cluster_map.png
│ ├── 04_timeline_flag_adoptions.png
│ ├── 05_research_summary_infographic.png
│ └── 06_interactive_summary.html
├── requirements.txt
├── LICENSE
└── README.md
- Python 3.8+
- Jupyter Notebook or JupyterLab
# Clone the repository
git clone https://github.com/Mariechanne/African-Flag-Color-Analysis
cd african_flags_project
# Create and activate a virtual environment (recommended)
python -m venv venv
source venv/bin/activate # macOS/Linux
venv\Scripts\activate # Windows
# Install dependencies
pip install -r requirements.txt
# Install geopandas separately (required for Notebook 04 map visualization)
pip install geopandasNote on geopandas: Due to complex binary dependencies,
geopandasis not included inrequirements.txt. On Windows, the easiest install path is via conda:conda install -c conda-forge geopandas
jupyter notebookRun notebooks in order:
| Step | Notebook | Output |
|---|---|---|
| 1 | 01_data_collection.ipynb |
african_countries_enhanced.csv |
| 2 | 02_color_extraction.ipynb |
african_flags_with_colors.csv |
| 3 | 03_color_similarity_clustering.ipynb |
african_flags_clustered.csv |
| 4 | 04_visualizations.ipynb |
All 6 visualizations in data/visualizations/ |
| Category | Libraries |
|---|---|
| Data Analysis | pandas, numpy |
| Machine Learning | scikit-learn (K-Means, Hierarchical Clustering) |
| Visualization | matplotlib, seaborn |
| Geospatial | geopandas (Natural Earth data) |
| Image Processing | Pillow |
| Data Collection | requests, beautifulsoup4 |
| Interactive Output | HTML5, CSS3, Vanilla JavaScript |
| Environment | jupyter, ipykernel |
| Metric | Value |
|---|---|
| Countries analyzed | 54 |
| Features per country | 39 |
| Total colors extracted | ~200 |
| Optimal clusters (k) | 7 |
| Silhouette score | 0.260 |
| Largest cluster | Cluster 4 — 26 countries (Panafricanist) |
| Smallest cluster | Cluster 7 — 1 country (Tunisia) |
| Flag images | 54 PNGs |
| Raw data files | 9 CSVs |
| Output visualizations | 5 static + 1 interactive |
Colonial heritage shows moderate correlation (50%) while geographic proximity shows strong correlation (90%). Neighboring countries share aesthetics regardless of which European power once governed them.
Both Panafricanist (Cluster 4) and Socialist liberation (Cluster 6) movements deliberately adopted specific color palettes as political statements — creating visual solidarity that cuts across colonial and geographic lines.
Several countries significantly redesigned their flags after independence: South Africa (1910 → 1994), Libya (1951 → 2011), Malawi (1964 → 2012). The analysis uses flag adoption year — not independence year — for temporal accuracy.
Islamic motifs (crescents, green) appear in North and East African flags but integrate with rather than override regional and political factors. Tunisia is the clearest case where religious-aesthetic continuity trumped continental solidarity.
- LAB color space — perceptual color distance for more human-accurate clustering
- Symbol-based features — stars, crescents, animals, shields as clustering dimensions
- Multi-modal clustering — combining color + layout structure + symbolism
- Continental comparison — same methodology applied to South American, Caribbean, or Pacific flags
- Longitudinal analysis — track flag redesigns over time as political regimes change
This project is licensed under the MIT License — see the LICENSE file for details.
Flag images sourced from flagcdn.com. Geographic data from Natural Earth.
This project was created as a data science portfolio piece demonstrating:
- Statistical methodology — feature engineering, unsupervised clustering, silhouette analysis
- Cultural awareness — historical context grounds every quantitative result
- Visual storytelling — 5 publication-quality charts + a fully interactive web report
- Technical range — data collection → processing → ML → static viz → interactive HTML
Author: Marie Chandeste MEDETADJI MIGAN Date: March 2026 Contact: Available for data science roles and collaborations
"Post-independence Africa chose its own colors. Flags tell a story of regional solidarity, ideological movements, and cultural assertion — more than they tell a story of colonial inheritance."