API: ML & importance

`MLModels`

NEExT.ml_models.MLModels trains and evaluates a model on graph embeddings. The framework’s train_ml_model wraps it; use the class directly for extra control over model_name and test_size.

MLModels

MLModels(
  graph_collection,
  embeddings,
  model_type="classifier",      # "classifier" | "regressor"
  model_name="xgboost",         # "xgboost" | "random_forest"
  balance_dataset=False,
  compute_feature_importance=False,
  sample_size=5,
  test_size=0.3,
  random_state=42,
  n_jobs=-1,
  parallel_backend="process",   # "process" | "thread"
)

compute() -> dict                 # metric lists, e.g. {"accuracy": [...]} or {"rmse": [...]}

XGBoost is used when available, otherwise scikit-learn’s random forest. See ML models.

`FeatureImportance`

NEExT.ml_models.FeatureImportance ranks structural features. The framework’s compute_feature_importance wraps it (passing n_iterations as the internal sample_size).

FeatureImportance

FeatureImportance(
  graph_collection,
  features,
  algorithm,                    # "supervised_greedy" | "supervised_fast" | "unsupervised"
  embedding_algorithm="approx_wasserstein",
  random_state=42,
  n_iterations=5,
  n_jobs=-1,
  parallel_backend="process",
)

compute() -> pandas.DataFrame

See Feature importance.