rankeval.scoring package¶
The rankeval.scoring
module includes utilities to score a model on a given dataset.
-
class
rankeval.scoring.
Scorer
(model, dataset)[source]¶ Bases:
object
Class for efficient scoring of an ensemble-based model composed of binary regression trees on a given dataset.
This class can be used for simple or detailed scoring, depending on the mode selected at scoring time. The document scores are cached as to avoid useless re-scoring. Thus, calling multiple times the score method does not involve the scoring activity to be executed again, except for a detailed scoring following a basic scoring. Indeed in this situation the scoring has to be repeated as to analyze in depth the scoring behaviour.
- model: RTEnsemble
- The model to use for scoring
- dataset: Dataset
- The dataset to use for scoring
- model : RTEnsemble
- The model to use for scoring
- dataset : Dataset
- The dataset to use for scoring
- y_pred : numpy array of float
- The predicted scores produced by the given model for each sample of the given dataset X
- partial_y_pred : numpy 2d-array of float
- The predicted score of each tree of the model for each dataset instance
-
get_partial_predicted_scores
()[source]¶ Provide an accessor to the partial scores produced by the given model for each sample of the given dataset X. Each partial score reflects the score produced by a single tree of the ensemble model to a single dataset instance. Thus, the returned numpy matrix has a shape of (n_instances, n_trees). The partial scores does not take into account the tree weights, thus for producing the final score is needed to multiply each row for the tree weight vector.
- scores : numpy 2d-array of float
- The predicted score of each tree of the model for each dataset instance
Submodules¶
rankeval.scoring.scorer module¶
Class for efficient scoring of an ensemble-based model composed of binary regression trees on a given dataset.
-
class
rankeval.scoring.scorer.
Scorer
(model, dataset)[source]¶ Bases:
object
Class for efficient scoring of an ensemble-based model composed of binary regression trees on a given dataset.
This class can be used for simple or detailed scoring, depending on the mode selected at scoring time. The document scores are cached as to avoid useless re-scoring. Thus, calling multiple times the score method does not involve the scoring activity to be executed again, except for a detailed scoring following a basic scoring. Indeed in this situation the scoring has to be repeated as to analyze in depth the scoring behaviour.
- model: RTEnsemble
- The model to use for scoring
- dataset: Dataset
- The dataset to use for scoring
- model : RTEnsemble
- The model to use for scoring
- dataset : Dataset
- The dataset to use for scoring
- y_pred : numpy array of float
- The predicted scores produced by the given model for each sample of the given dataset X
- partial_y_pred : numpy 2d-array of float
- The predicted score of each tree of the model for each dataset instance
-
get_partial_predicted_scores
()[source]¶ Provide an accessor to the partial scores produced by the given model for each sample of the given dataset X. Each partial score reflects the score produced by a single tree of the ensemble model to a single dataset instance. Thus, the returned numpy matrix has a shape of (n_instances, n_trees). The partial scores does not take into account the tree weights, thus for producing the final score is needed to multiply each row for the tree weight vector.
- scores : numpy 2d-array of float
- The predicted score of each tree of the model for each dataset instance