Unsupervised Clustering — K-Means + Streamlit Live Demo

An end-to-end unsupervised learning project that applies K-Means clustering to a tabular dataset and deploys an interactive Streamlit demo for exploring clusters, tuning parameters, and visualizing results.

Live Demo: https://clustering-segmentation-explorer-nadine-shill.streamlit.app/
Repo: https://github.com/nadine-ramirez/unsupervised-clustering-kmeans
Notebook: https://github.com/nadine-ramirez/unsupervised-clustering-kmeans/blob/main/Clustering.ipynb

---

Overview

Clustering is a core unsupervised learning technique used to discover structure in data when labels aren’t available. This project focuses on a practical workflow:

• Load and preprocess structured data
• Fit a K-Means model
• Evaluate cluster quality (e.g., inertia / silhouette score)
• Visualize clusters (2D via PCA)
• Deploy a Streamlit app to explore results interactively

---

Streamlit App Features

The Streamlit app (app.py) lets you:

• Explore a built-in sample dataset (Iris) or use your own dataset (optional if implemented)
• Select clustering features
• Choose the number of clusters (K) with a slider
• View:
  • Cluster assignments
  • PCA visualization of clusters
  • Summary statistics by cluster
  • Basic clustering metrics (inertia, silhouette score)
• Export results as a CSV (optional if implemented)

---

How It Works (High Level)

1. Preprocessing
   • Numeric feature selection
   • Scaling (if enabled)
2. Model
   • K-Means clustering using scikit-learn
3. Dimensionality Reduction
   • PCA to project features into 2D for visualization
4. Outputs
   • Cluster labels added to the dataset
   • Visual + metric feedback to help choose a good K

---

Project Structure