DBSCAN Versus k-means and Hierarchical Clustering

Now that you've reached an understanding of how DBSCAN is implemented and how many different hyperparameters you can tweak to drive performance, let's survey how it compares to the clustering methods we covered in Chapter 1, Introduction to Clustering and Chapter 2, Hierarchical Clustering.

You may have noticed in Activity 5, Comparing DBSCAN with k-means and Hierarchical Clustering, that DBSCAN can be a bit finnicky when it comes to finding the optimal clusters via silhouette score. This is a downside of the neighborhood approach – k-means and hierarchical clustering really excel when you have some idea regarding the number of clusters in your data. In most cases, this number is low enough that you can iteratively try a few different numbers and see how it performs. DBSCAN, instead, takes a more bottom-up approach by working with your hyperparameters and finding the clusters it views as important. In practice, it is helpful to consider DBSCAN...