graphtester.recommend

graphtester.recommend(dataset: Dataset, metrics: List[str], max_feature_count=3, features_to_test=None, node_features=True, edge_features=True, ignore_original_features=False, fast=True, iterations=1) → RecommendationResult

Recommend additional features (methods) to add to a dataset.

Test all possible features and recommend the ones that add the most 1-WL-efficiency to the dataset. Features are tested by adding them to the dataset one at a time and evaluating the dataset. The feature that adds the most efficiency is then added to the dataset and the process is repeated until all features have been tested, or full efficiency has been reached in all dimensions.

Parameters:

datasetDataset

The dataset to recommend features for.

metricsList[str]

The metrics to evaluate the dataset on.

max_feature_countint, optional

The maximum number of features to combine into a set, by default 3

features_to_testList[str], optional

The features to test, by default None. If None, all features will be tested, depending on node_features, edge_features and fast arguments (see below).

node_featuresbool, optional

Whether to recommend node features, by default True

edge_featuresbool, optional

Whether to recommend edge features, by default True

ignore_original_featuresbool, optional

Whether to ignore the original features of the dataset, by default False

fastbool, optional

Whether to only use the features that are scalable to large datasets, by default True. Ignored if features_to_test is not None.

iterationsint, optional

The number of iterations to run the comparison for, by default 1

Returns:

RecommendationResult

The recommendation result.