Unsupervised Learning on the Flags Dataset: PCA and K-Means Clustering

This project explores how Principal Component Analysis (PCA) and K-Means clustering perform on the Flags dataset from the UCI Machine Learning Repository. The study highlights how datasets with mixed feature types influence interpretability and model performance.

Project Structure

• flags_pca_clustering.ipynb — Main Jupyter notebook containing code and explanations.
• flags_pca_clustering.html — Exported HTML version of the notebook.
• flags_pca_clustering.pdf — Exported PDF version of the notebook.
• figures/ — Contains exported plots mainly for reference; all key figures are already embedded in the outputs.

Key Points

• Perform PCA and K-Means clustering on the Flags dataset.
• Conduct exploratory data analysis to visualize trends and correlations.
• Assess PCA decomposition onto first two principal components.
• Analyze clustering results and discuss the tradeoff between parsimony and interpretability.

Requirements

• Python (3.10.16 recommended)
• Jupyter Notebook / Jupyter Lab
• Python packages: pandas, numpy, matplotlib, seaborn, altair, scikit-learn, ucimlrepo

You can install the required packages using:

pip install pandas numpy matplotlib seaborn altair scikit-learn ucimlrepo

How to Use

1. Clone or download this repository.
2. Open flags_pca_clustering.ipynb in Jupyter Notebook or Jupyter Lab and load the dataset via the ucimlrepo package.
3. Run all cells to reproduce results, figures, and exported HTML/PDF outputs.