Node-level learning

Node-level learning is the task of learning and predicting attributes or behaviors of individual nodes in a graph. Depending on the type and range of the target variable, node-level learning can be categorized into four subtasks: node classification, node regression, node clustering, and node anomaly detection.

Node classification

Node classification in graphs is an ML task that aims to assign labels or categories to nodes based on their features and connections. For example, in an e-commerce graph, we can classify users into different preference groups based on their interactions with items. To do this, we need to extract features from both nodes and edges, such as demographic information, item attributes, popularity, reviews, frequency of purchases, and time spent on items. These features capture the characteristics and preferences of users and items, as well as the strength and nature of their interactions. By training an ML model on a subset of labeled...