Unsupervised Learning - Using the Outcome of Learning

Question

My understanding in Unsupervised Learning is that -- when you want a computer to learn on its own by examining a large dataset. The goal is to establish some form of cluster or association-based grouping of data from the given input aka "Find patterns - known/unknown"

So my question is - "Now what do we do with it?" - e.g. I have now identified a pattern using Unsupervised Learning. Do I now look for Supervised Learning to empower my Data Insights work? Or, do I make decision as a human/team/committee to do certain things based on those?

Answer 1

Unsupervised Learning is a collection of tools and you can use those tools for many purposes.

Sometime, it's just for data visualisation / data exploration (see UMAP for example), to better understand the problem and get the idea how to tackle it. Sometimes it will lead to business decision like building processes for a specific cluster or building submodels for specific clusters.

Sometime it helps deal with sub-problems of data-science, like outliers identification (see Hdbscan or isolation forests for example). What to do with those outliers depends on your business case.

Sometimes it helps define intermediate 'targets' to perform semi-supervised or supervised learning, like using some self-organisaed map, identify a cluster and use that cluster as a target. It help when you are not sure you have identified all your positive targets or when you don't wnat to label everything.

Answer 2

You can do both, meaning

a) look for Supervised Learning to empower my Data Insights

b) make decision as a human/team/committee to do certain things based on those

What does that mean?

in the case of a) you can think of unsupervised learning as representation learning-meaning you learn the best quantitve represenation of your data, and than right at the end, i.e. for example last layer you add an output neuron to produce binary classification. Than you retrain (you need labels for this obviously) your "unsupervised learning network" to adjust it to make classification

b) You analyse the clusters of the unsupervised task for example. Where you build a system to order the input data into these clusters and you can analyse the relationships (this would not require labels, and this is obviously just one example of unsupervised approach)