CandidateRankingModel

class rectools.models.ranking.candidate_ranking.CandidateRankingModel(candidate_generators: List[CandidateGenerator], splitter: Splitter, reranker: Reranker, sampler: Optional[NegativeSamplerBase] = None, feature_collector: Optional[CandidateFeatureCollector] = None, verbose: int = 0)

Bases: ModelBase

Candidate Ranking Model for recommendation systems.

Inherited-members

Parameters

• candidate_generators (List[CandidateGenerator]) –

• splitter (Splitter) –

• reranker (Reranker) –

• sampler (Optional[NegativeSamplerBase]) –

• feature_collector (Optional[CandidateFeatureCollector]) –

• verbose (int) –

Methods

dumps()	Serialize model to bytes.
fit(dataset, *args, **kwargs)	Fit model.
fit_partial(dataset, *args, **kwargs)	Fit model.
from_config(config)	Create model from config.
from_params(params[, sep])	Create model from parameters.
get_config([mode, simple_types])	Return model config.
get_full_candidates_with_targets(...)	Prepare candidates with target values set from first-stage candidate generators.
get_params([simple_types, sep])	Return model parameters.
get_train_with_targets_for_reranker(dataset)	Prepare training data for the reranker.
load(f)	Load model from file.
loads(data)	Load model from bytes.
recommend(users, dataset, k, filter_viewed)	Generate k recommendations for specified users using the dataset.
recommend_to_items(target_items, dataset, k)	Recommend items for target items.
save(f)	Save model to file.
split_to_history_dataset_and_train_targets(...)	Split interactions into history and train sets for first-stage and second-stage model training.

Attributes

recommends_for_cold
recommends_for_warm
require_recommend_context	Indicates whether recommendation context is required for predictions.

get_full_candidates_with_targets(train_targets: DataFrame, history_dataset: Dataset) → DataFrame

Prepare candidates with target values set from first-stage candidate generators.

Parameters

• train_targets (pd.DataFrame) – DataFrame containing training targets.

• history_dataset (Dataset) – The dataset to fit the candidate generators on.

Returns

DataFrame with target values set.

Return type

pd.DataFrame

get_train_with_targets_for_reranker(dataset: Dataset) → DataFrame

Prepare training data for the reranker.

Parameters

dataset (Dataset) – The dataset to prepare training data from.

Returns

DataFrame containing training data with targets and 2 extra columns: Columns.User, Columns.Item.

Return type

pd.DataFrame

recommend(users: Union[Sequence[Hashable], ndarray], dataset: Dataset, k: int, filter_viewed: bool, items_to_recommend: Optional[Union[Sequence[Hashable], ndarray]] = None, add_rank_col: bool = True, on_unsupported_targets: Literal['ignore', 'warn', 'raise'] = 'raise', context: Optional[DataFrame] = None, force_fit_candidate_generators: bool = False) → DataFrame

Generate k recommendations for specified users using the dataset.

Parameters

• users (ExternalIds) – List of user ids for whom recommendations are generated.

• dataset (Dataset) – Dataset containing user-item interaction data and possibly additional features.

• k (int) – The number of recommendations to generate for each user.

• filter_viewed (bool) – If true, viewed items will be excluded from the recommendations.

• items_to_recommend (ExternalIds, optional, default None) – List of item ids from which recommendations should be generated. If not provided, it will include all items available in the dataset.

• add_rank_col (bool, default True) – If true, a rank column is added to the returned DataFrame. The rank column shows the position of the item in the sorted order of predictions.

• on_unsupported_targets (ErrorBehaviour, default "raise") – Controls the behavior when a target is encountered during prediction, for which the Model makes no prediction. If “raise”, a ValueError is raised. If “warn”, it outputs a warning, and if “ignore”, it silently continues.

• context (pd.DataFrame, optional, default None) – Is not supported for this model. If provided, it will be ignored.

• force_fit_candidate_generators (bool, default False) – If true, the candidate generators are fitted even if they are already fitted.

Returns

DataFrame with the recommended items for users.

Return type

pd.DataFrame

split_to_history_dataset_and_train_targets(dataset: Dataset, splitter: Splitter) → Tuple[Dataset, DataFrame, Dict[str, Any]]

Split interactions into history and train sets for first-stage and second-stage model training.

Parameters

• dataset (Dataset) – The dataset to split.

• splitter (Splitter) – The splitter to use for splitting the dataset.

Returns

Tuple containing the history dataset, train targets, and fold information.

Return type

Dataset, pd.DataFrame, dict(str -> any)