Unsupervised learning - Clustering and dimensionality reduction

 

Unsupervised learning can be a useful tool for finding relationships or grouping that may exist with large data sets. This can be extremely useful for finding deeper relationships than what may exist from just looking at correlation relationships. It can also be useful at finding links between a large set of assets and exogenous factors. More finance work has been done using PCA as a way of generating simple dimensional reductions. It is an east way to eliminate a primary common factor across stocks. 

I have been using unsupervised learning to help better gain diversification within a portfolio of futures markets especially within the commodity space.