Why Graph-modeling Frameworks are the Future of Unsupervised Learning

An iterative graph-modeling methodology for estimating feature importance in unsupervised learning scenarios

Co-authored by Abhishek Singh, Machine Learning Engineer at Bayer Pharmaceuticals, former Microsoft, JPMorgan Chase & Co, HSBC, and by Cristiana de Azevedo von Stosch, Digital Health Data Science at Bayer Pharmaceuticals.

Iteration of the directed graph methodology on the Titanic dataset. Image by author.

Not all features are created equal. Moreover, feature determination importance is still a fundamental problem in machine learning.

Most methods of …

clustering feature-importance frameworks future graph learning modeling unsupervised unsupervised learning