Summary

In this chapter, we considered several use cases of ML algorithms on network datasets. This included UL on a friendship network through fitting k-means and spectral clustering. We considered k-means clustering on both the original dataset of activities in which individuals participated and the original dataset, with added network metrics to improve clustering accuracy. We then turned to SL and SSL on networks and collections of networks through a type of DL algorithm called GNNs. We accurately predicted the labels of individuals in Zachary’s Karate Network dataset through a shallow GNN and compared results with other existing solutions to this network classification problem. In Chapter 10, we'll mine educational data for causal relationships using network tools related to conditional probability.