Unsupervised Machine Learning on Network Data
Welcome to another exciting chapter exploring network science and data science together. In the last chapter, we used supervised ML to train a model that was able to detect the revolutionaries from the book Les Miserables, using graph features alone. In this chapter, we are going to explore unsupervised ML and how it can also be useful in graph analysis as well as node classification with supervised ML.
The order these two chapters have been written in was intentional. I wanted you to learn how to create your own training data using graphs rather than being reliant on embeddings from unsupervised ML. The reason for this is important: when you rely on embeddings, you lose the ability to interpret why ML models have been classified the way that they have. You lose interpretability and explainability. The classifier essentially works as a black box, no matter which model you use. I wanted to show you the interpretable and explainable approach...