VotingChainladder

VotingChainladder#

class chainladder.VotingChainladder(estimators, *, weights=None, default_weighting=None, n_jobs=None, verbose=False)[source]#

Prediction voting chainladder method for unfitted estimators.

A voting chainladder is an ensemble meta-estimator that fits several base chainladder methods, each on the whole triangle. Then it combines the individual predictions based on a matrix of weights to form a final prediction.

Read more in the User Guide.

Added in version 0.8.0.

Parameters:

estimators: list of (str, estimator) tuples: Invoking the fit method on the VotingChainladder will fit clones of those original estimators that will be stored in the class attribute self.estimators_. An estimator can be set to 'drop' using set_params.
weights: array callable or dict, default=None: array: Numpy array of shape (index, columns, origin, n_estimators). Minimum shape required is (origin, n_estimators). Lower dimensional weight arrays will have missing dimensions repeated to match the shape of the triangle. list: List of weights where each weight is a list of length n_estimators. dict: A dictionary where the origin is mapped to a weighting tuple. Missing origin periods will be given default_weighting. callable: A callable that returns weighting tuples. None uses default_weighting.
default_weighting: tuple of shape (n_estimators, ), default=None: Default weighting to use where a weight was not provided or if weights is None. None uses a typle of all ones which is equivalent to averaging the predictions of the estimators.
n_jobs: int, default=None: The number of jobs to run in parallel for fit. None means 1 unless in a joblib.parallel_backend context. -1 means using all processors. for more details.
verbose: bool, default=False: If True, the time elapsed while fitting will be printed as it is completed.

Attributes:

estimators_: list of chainladder estimators: The collection of fitted sub-estimators as defined in estimators that are not ‘drop’.
named_estimators_: Bunch: Attribute to access any fitted sub-estimators by name.

Examples

An actuary reserving the RAA excess casualty book leans on the development-based Chainladder method for the mature accident years but shifts toward the more stable exposure-based methods in the recent, least developed years. The weights matrix has one row per accident year and one column per estimator, phasing the blend from Chainladder to BornhuetterFerguson and CapeCod.

import numpy as np

raa = cl.load_sample('RAA')
cl_ult = cl.Chainladder().fit(raa).ultimate_  # Chainladder Ultimate
apriori = cl_ult * 0 + (float(cl_ult.sum()) / 10)  # Mean Chainladder Ultimate
bcl = cl.Chainladder()
bf = cl.BornhuetterFerguson()
cc = cl.CapeCod()
estimators = [('bcl', bcl), ('bf', bf), ('cc', cc)]
weights = np.array([[0.6, 0.2, 0.2]] * 4 + [[0, 0.5, 0.5]] * 3 + [[0, 0, 1]] * 3)
vot = cl.VotingChainladder(estimators=estimators, weights=weights)
vot.fit(raa, sample_weight=apriori)
print(vot.ultimate_)

              2261
18834.000000
16875.500226
24058.534810
28542.580970
28236.843134
19905.317262
18947.245455
23106.943030
20004.502125
21605.832631

Building on the example above and reusing its estimators, raa and apriori, the actuary can instead blend all three methods in every accident year. Omitting weights applies default_weighting to each accident year; here Chainladder receives half of the total weight.

blended = cl.VotingChainladder(
    estimators=estimators, default_weighting=(2, 1, 1)
)
blended.fit(raa, sample_weight=apriori)
print(blended.ultimate_)

              2261
18834.000000
16879.886803
24052.325782
28502.440672
28581.789739
19703.210223
18348.274023
23483.819232
17908.906366
19849.185129

fit(X, y=None, sample_weight=None)[source]#

Fit the estimators.

Parameters:

XTriangle: Loss data to which the voting will be applied.
yNone: Ignored
sample_weightTriangle, default=None: Exposure to be used in the calculation. Required if any of the estimators are exposure based.

Returns:

selfobject: Fitted estimator.

fit_transform(X, y=None, sample_weight=None)[source]#

Fit and return predictions for VotingChainladder

Parameters:

XTriangle: Loss data to which the model will be applied.
yNone: Ignored
sample_weightTriangle, default=None: Exposure to be used in the calculation. Required if any of the estimators are exposure based.

Returns:

X_new: Triangle: Loss data with VotingChainladder ultimate applied

predict(X, sample_weight=None)[source]#

Predicts the voting chainladder ultimate on a new triangle X

Predicts the ultimate for each of the estimators and combines them into a single ultimate based on the weights given.

Parameters:

XTriangle: Loss data to which the model will be applied.
sample_weightTriangle, default=None: Exposure to be used in the calculation. Required if any of the estimators are exposure based.

Returns:

X_new: Triangle: Loss data with VotingChainladder ultimate applied

transform(X, sample_weight=None)[source]#

Return predictions for VotingChainladder

Parameters:

XTriangle: Loss data to which the model will be applied.
sample_weightTriangle, default=None: Exposure to be used in the calculation. Required if any of the estimators are exposure based.

Returns:

X_new: Triangle: Loss data with VotingChainladder ultimate applied

Inherited Methods

`VotingChainladder.fit_predict`
`VotingChainladder.get_metadata_routing`	Get metadata routing of this object.
`VotingChainladder.get_params`	Get the parameters of an estimator from the ensemble.
`VotingChainladder.intersection`	Given two Triangles with mismatched indices, this method aligns their indices
`VotingChainladder.pipe`	Apply `func(self, args, *kwargs)`.
`VotingChainladder.set_backend`	Converts triangle array_backend.
`VotingChainladder.set_output`	Set output container.
`VotingChainladder.set_params`	Set the parameters of an estimator from the ensemble.
`VotingChainladder.to_json`	Serializes triangle object to json format
`VotingChainladder.to_pickle`	Serializes triangle object to pickle.
`VotingChainladder.validate_X`
`VotingChainladder.validate_weight`

VotingChainladder

Contents

VotingChainladder#