Java: Data Science Made Easy

Extracting features from the graph

We need to extract some features that we will put to the Machine Learning model for training. For this dataset, all the information we have is the graph itself and nothing more: we do not have any external information such as author's affiliation. Of course, if we had it, it would be no problem to add it to the model. So let's discuss which features we can extract from the graph alone.

For graph models, there can be two kinds of features: node features (authors) and edge features (the coauthorship relation).

There are many possible features we can extract from graph nodes. For example, among others, we can consider the following:

• Degree: This is the number of coauthors this author has.
• Page Rank: This is ...