18 Clustering based on density: DBSCAN and OPTICS

Our penultimate stop in unsupervised learning techniques brings us to density-based clustering. Density-based clustering algorithms aim to achieve the same thing as k-means and hierarchical clustering: to partition a dataset into a finite set of clusters that reveal a grouping structure in our data.

18.1  What is density-based clustering?

18.1.1  How does the DBSCAN algorithm learn?

18.1.2  How does the OPTICS algorithm learn?

18.2  Building our first DBSCAN model

18.2.1  Loading and exploring the banknote dataset

18.2.2  Tuning the epsilon and minPts hyperparameters

18.3  Building our first OPTICS model

18.4  Strengths and weaknesses of density-based clustering

18.5  Summary

18.6  Solutions to exercises