tempor.methods.time_to_event.helper_embedding module¶

Helper class for embedding time-series data for time-to-event analysis using Dynamic DeepHit embeddings.

class tempor.methods.time_to_event.helper_embedding.OutputTimeToEventAnalysis[source]¶

Bases: object

Helper base class for time-to-event analysis models.

abstract fit(X: DataFrame, T: Series, Y: Series) → Self[source]¶

Fit the model.

Parameters:¶

X : pd.DataFrame¶: Input covariates of fixed shape - time-series data should be embedded first.
T : pd.Series¶: Event times.
Y : pd.Series¶: Event values.

Returns:¶

Fitted model.

Return type:¶

Self

abstract predict_risk(X: DataFrame, time_horizons: list) → DataFrame[source]¶

Predict risk scores.

Parameters:¶

X : pd.DataFrame¶: Input covariates of fixed shape - time-series data should be embedded first.
time_horizons : List¶: Time horizons to predict risk at.

Returns:¶

Risk scores.

Return type:¶

pd.DataFrame

class tempor.methods.time_to_event.helper_embedding.DDHEmbedding(emb_model: DynamicDeepHitModel)[source]¶

Bases: object

Survival analysis embedding creation for time-series with tempor.models.ddh.DynamicDeepHitModel.

Parameters:¶

emb_model : DynamicDeepHitModel¶: tempor.models.ddh.DynamicDeepHitModel to use for temporal feature embedding.

prepare_fit(data: BaseDataset) → tuple[ndarray, ndarray, ndarray][source]¶

Prepare data for fitting.

Parameters:¶

data : dataset.BaseDataset¶: Input dataset.

Returns:¶

Processed covariate data, event times, event values.

Return type:¶

Tuple[np.ndarray, np.ndarray, np.ndarray]

prepare_predict(data: PredictiveDataset, horizons: list[float] | list[int] | list[Timestamp], *args: Any, **kwargs: Any) → ndarray[source]¶

Prepare data for prediction.

Parameters:¶

data : dataset.PredictiveDataset¶: Input dataset.
horizons : data_typing.TimeIndex¶: Time horizons to predict risk at.
*args : Any: Additional arguments.
**kwargs : Any: Additional keyword arguments.

Returns:¶

Processed data.

Return type:¶

np.ndarray

static hyperparameter_space(*args: Any, **kwargs: Any) → list[Params][source]¶

class tempor.methods.time_to_event.helper_embedding.DDHEmbeddingTimeToEventAnalysis(output_model: OutputTimeToEventAnalysis, emb_model: DynamicDeepHitModel)[source]¶

Bases: DDHEmbedding

Survival analysis embedding creation for time-series with tempor.models.ddh.DynamicDeepHitModel followed by output_model OutputTimeToEventAnalysis survival analysis estimator.

Parameters:¶

output_model : OutputTimeToEventAnalysis¶: Output model to use for predicting risk.
emb_model : DynamicDeepHitModel¶: tempor.models.ddh.DynamicDeepHitModel to use for temporal feature embedding.

fit(data: BaseDataset, *args: Any, **kwargs: Any) → Self[source]¶

Fit the model.

Parameters:¶

data : dataset.BaseDataset¶: Input dataset.
*args : Any: Additional arguments.
**kwargs : Any: Additional keyword arguments.

Returns:¶

Fitted model.

Return type:¶

Self

predict(data: PredictiveDataset, horizons: list[float] | list[int] | list[Timestamp], *args: Any, **kwargs: Any) → TimeSeriesSamplesBase[source]¶

Predict risk scores.

*args and **kwargs will be passed to self.emb_model.predict_emb(). E.g. batch_size batch size parameter can be provided this way.

Parameters:¶

data : dataset.PredictiveDataset¶: Input dataset.
horizons : data_typing.TimeIndex¶: Time horizons to predict risk at.
*args : Any: Additional arguments. Passed to self.emb_model.predict_emb()
**kwargs : Any: Additional keyword arguments. Passed to self.emb_model.predict_emb()

Returns:¶

Predicted risk scores.

Return type:¶

samples.TimeSeriesSamplesBase

static hyperparameter_space(*args: Any, **kwargs: Any) → list[Params][source]¶