lkauto.explicit package¶
Submodules¶
lkauto.explicit.explicit_evaler module¶
-
class
lkauto.explicit.explicit_evaler.ExplicitEvaler(train: pandas.core.frame.DataFrame, optimization_metric, filer: lkauto.utils.filer.Filer, validation=None, random_state=42, split_folds: int = 1, split_strategie: str = 'user_based', split_frac: float = 0.25, ensemble_size: int = 50, minimize_error_metric_val: bool = True)¶ Bases:
objectthe ExplicitEvaler class handles the evaluation of rating prediction models. An Evaluation run consists of training a model and to predict and evaluate the performance on a validation split.
-
train¶ pandas dataset containing the train split.
- Type
pd.DataFrame
-
optimization_metric¶ LensKit prediction accuracy metric used to evaluate the model (either rmse or mae)
- Type
function
-
validation¶ pandas dataset containing the validation split.
- Type
pd.DataFrame
-
random_state¶ The random number generator or seed (see
lenskit.util.rng()).
-
split_folds¶ The number of folds of the validation split
-
split_strategie¶ The strategie used to split the data. Possible values are ‘user_based’ and ‘row_based’
-
split_frac¶ The fraction of the data used for the validation split. If the split_folds value is greater than 1, this value is ignored.
-
ensemble_size¶ The number of models used to build the final ensemble predictor.
-
minimize_error_metric_val¶ If True, the error metric is minimized. If False, the error metric is maximized. This parameter needs to be set in corelation with the optimization metric.
-
evaluate_explicit(config_space: ConfigurationSpace) → float¶
-
evaluate(config_space: ConfigSpace.configuration_space.ConfigurationSpace) → float¶ evaluates model defined in config_space
The config_space parameter defines a model. This model is build, trained and evaluated with the validation split.
- Parameters
config_space (ConfigurationSpace) – configuration space containing information to build a model
- Returns
validation_error – the error of the considered model
- Return type
-