Refactor K-Means clustering to use Wine Quality dataset

[kaleb-kebede]

Summary

This PR updates the Unsupervised Learning assignment to use the wine-quality-white-and-red.csv dataset. The previous medicineData.csv contained only 4 rows, which was insufficient for training a valid K-Means model.

Key Changes

• Dataset Switch: Changed source URL to the Wine Quality dataset (~6,000 rows) to generate meaningful clusters.
• Preprocessing: * Added duplicate removal.
  • Implemented StandardScaler to normalize feature variance (crucial for K-Means).
• Model Tuning:
  • Generated a valid Elbow Method graph to determine the optimal number of clusters ($k$).
  • Selected optimal_k = 4 based on the inertia plot.
• Visualization: * Used PCA to reduce dimensionality to 2D for plotting.
  • Added centroids to the scatter plot for clarity.

Results

The model now successfully segments wine samples into distinct clusters and provides a valid prediction for new input data.