Link Prediction Classifier

Feature Extraction

The function create_features extracts various features between pairs of nodes in a graph G using network metrics and node attributes.

For each pair of nodes:

• Common neighbors: Number of nodes that are neighbors to both nodes.
• Jaccard coefficient: A measure of similarity based on shared neighbors.
• Preferential attachment: A score that measures the likelihood of a link based on node degrees.
• Adamic-Adar index: Another similarity measure based on shared neighbors, weighted by their degrees.
• Same attribute: A binary feature indicating whether the two nodes have the same attribute value.
• Degree difference: Absolute difference in degrees of the two nodes.
• Total degree: Sum of the degrees of the two nodes.
• Average clustering coefficient: The average clustering coefficient of the two nodes.
• Same community: A binary feature indicating if the two nodes belong to the same community based on label propagation.

Random Forest Model and Hyperparameter Tuning

• Randomised search with cross-validation (RandomizedSearchCV) to tune the model. The model is optimized based on the F1 score.

Model Evaluation

The best Random Forest model is evaluated on the test set, where accuracy, precision, recall, and F1 score are printed. Feature importance scores are also displayed.