Java: Data Science Made Easy

Supervised learning via clustering

Unsupervised learning can be used as a model for supervised learning, and depending on the supervised problem we have, it can be either classification via clustering or regression via clustering.

This approach is relatively straightforward. First, you associate each item with some cluster ID, and then:

• For the binary classification problem, you output the probability of seeing the positive class in the cluster
• For regression, you output the mean value across the cluster

Let's look at how we can do this for regression. At the beginning, we run K-means on the original data as usual:

int k = 250; int maxIter = 10; int runs = 1; KMeans km = new KMeans(X, k, maxIter, runs);

Often it makes sense to pick a ...