Introduction to Density-Based Clustering (DBSCAN)

Density-based clustering or DBSCAN is one of the most intuitive forms of clustering. It is very easy to find naturally occurring clusters and outliers in data with this type of clustering. Also, it doesn't require you to define a number of clusters. For example, consider the following figure:

Figure 2.2: A sample scatter plot

There are four natural clusters in this dataset and a few outliers. So, DBSCAN will segregate the clusters and outliers, as depicted in the following figure, without you having to tell it how many clusters to identify in the dataset:

Figure 2.3: Clusters and outliers classified by DBSCAN

So, DBSCAN can find regions of high density separated by regions of low density in a scatter plot.

Steps for DBSCAN

As mentioned before, DBSCAN doesn't require you to choose a number of clusters, but you have to choose the other two parameters to perform DBSCAN. The first parameter is commonly denoted by ε (epsilon), which denotes the maximum...