Node Classification (Supervised)
Explore the fundamentals of supervised node classification in graph machine learning. Understand how to predict missing node labels using graph embeddings and machine learning models like support vector machines. Gain practical experience with real-world datasets, including Twitch UK and Zachary's Karate Club, to build and evaluate classification models that identify node labels accurately.
We'll cover the following...
Often, nodes in a graph come with some labels or tags. Some labels are usually missing when dealing with a vast graph database. Our task is to find or predict the missing labels using the knowledge from the available nodes. Yes, that's right! This sounds like a problem machine learning can solve.
Node classification is a task in which we build a predictive model, such as a binary classification task (in the case of two different types of labels) or a multiclass classification task (greater than two labels), to identify the missing cases.
Input features
The target class in the machine learning problem setting is the node labels. The predictor class/variable is the node embeddings. The input features’ dimensions are user-defined and can be treated as a hyperparameter. Any machine learning ...