Day Summary:
ML models developed: K-means, DBSCAN, HDBSCAN, and GMM (Gaussian Mixture Model)
Best results were HDBSCAN and GMM!!
Discussion
KMeans (4 clusters): Very crude, forces the clusters.
Old HDBSCAN (57 clusters): Too much detail, hard to explain.
GMM (30 clusters): Good balance, probabilistic clusters.
HDBSCAN (31 clusters): Large, clear, and meaningful groups. High noise, but fair, it only captures what’s truly atypical.