tick.survival.ConvSCCS

tick.survival.ConvSCCS(n_lags: numpy.array, penalized_features: numpy.array = None, C_tv=None, C_group_l1=None, step: float = None, tol: float = 1e-05, max_iter: int = 100, verbose: bool = False, print_every: int = 10, record_every: int = 10, random_state: int = None)[source]

ConvSCCS learner, estimates lagged features effect using TV and Group L1 penalties. These penalties constrain the coefficient groups modelling the lagged effects to ensure their regularity and sparsity.

Parameters

n_lags : numpy.ndarray, shape=(n_features,), dtype=”uint64”

Number of lags per feature. The model will regress labels on the last observed values of the features over their corresponding n_lags time intervals. n_lags values must be between 0 and n_intervals - 1.

penalized_features : numpy.ndarray, shape=(n_features,), dtype=”bool”, default=None

Booleans indicating whether the features should be penalised or not. If set to None, pernalize all features.

C_tv : float, default=None

Level of TV penalization TV penalization. This value should be None or greater than 0.

C_group_l1 : float, default=None

Level of group Lasso penalization. This value should be None or greater than 0.

step : float, default=None

Step-size parameter, the most important parameter of the solver. If set to None, it will be automatically tuned as step = 1 / model.get_lip_max().

tol : float, default=1e-5

The tolerance of the solver (iterations stop when the stopping criterion is below it).

max_iter : int, default=100

Maximum number of iterations of the solver, namely maximum number of epochs.

verbose : bool, default=False

If True, solver verboses history, otherwise nothing is displayed.

print_every : int, default=1

Print history information every time the iteration number is a multiple of print_every. Used only is verbose is True.

record_every : int, default=1

Save history information every time the iteration number is a multiple of record_every.

random_state : int, default=None

If not None, the seed of the random sampling.

Attributes

n_cases : int (read-only)

Number of samples with at least one outcome.

n_intervals : int (read-only)

Number of time intervals.

n_features : int (read-only)

Number of features.

n_coeffs : int (read-only)

Total number of coefficients of the model.

coeffs : list (read-only)

List containing 1-dimensional np.ndarray (dtype=float) containing the coefficients of the model. Each numpy array contains the (n_lags + 1) coefficients associated with a feature. Each coefficient of such arrays can be interpreted as the log relative intensity associated with this feature, k periods after exposure start, where k is the index of the coefficient in the array.

intensities : list (read-only)

List containing 1-dimensional np.ndarray (dtype=float) containing the intensities estimated by the model. Each numpy array contains the relative intensities of a feature. Element of these arrays can be interpreted as the relative intensity associated with a feature, k periods after exposure start, where k is the index of the coefficient in the array.

confidence_intervals : Confidence_intervals (read-only)

Coefficients refitted on the model and associated confidence intervals computed using parametric bootstrap. Refitted coefficients are projected on the support of the coefficients estimated by the penalised model. Refitted coefficients and their confidence intervals follow the same structure as coeffs.

References

Morel, M., Bacry, E., Gaïffas, S., Guilloux, A., & Leroy, F. (Submitted, 2018, January). ConvSCCS: convolutional self-controlled case series model for lagged adverse event detection