Julia Programming Projects

Choosing the optimal number of clusters for k-means (and other algorithms)

Depending on the nature of the data and the problem you'll be looking to solve, the number of clusters can come as a business requirement, or it may be an obvious choice (as in our case, where we wanted to identify low, middle, and high business density zones and so ended up with three clusters). However, in some cases, the answer might not be that obvious. In such situations, we'll need to apply a different algorithm to evaluate the optimal number of clusters.

One of the most common is the Elbow method. It is an iterative approach where we run the clustering algorithm with different values for k, for example between 1 and 10. The goal is to compare the total intra-cluster ...