Glossary#
This glossary hopes to definitively represent the tacit and explicit
conventions applied in chanladder and its API.
General Terms#
backend
The storage of the numerical representation of a [Triangle]{.title-ref}. It can be ‘numpy’ for a dense CPU bound representation, ‘sparse’ for a sparse matrix CPU bound representation, or ‘cupy’ for a dense GPU bound representation.
estimator
Any scikit-learn style class that can be
fitto Triangle data.
hyperparameter
An initial parameter of an estimator that can be set before the estimator is fit.
predictor
An estimator that has the
predictmethod. All IBNR estimators ofchainladderare predictors
transformer
An estimator that has the
transformmethod. The transform method returns instances of a Triangle. All estimators other than IBNR estimators are transformers.
Class API#
Triangle
The core data structure of the
chainladderpackage. It emulatespandas’s functionality.
Development
Transformers that produce, at a minimum, a multiplicative
ldf_property.
Tail
Transformers that extends the Development estimator properties beyond the edge of a Triangle.
IBNR
Predictors that produce, at a minimum, an
ultimate_,ibnr_,full_expectation_andfull_triangle_property.
Workflow
Meta-estimators that allow for composition of other estimators.
Adjustments
Estimators that allow for the adjustment of the values of a Triangle.
Triangle Concepts#
axis
Represents one of the four dimensions of a [Triangle]{.title-ref} instance. The four axes are
index,columns,originanddevelopment.valuationrepresents an additional axis implicit in the Triangle.
index
The first axis of a 4D Triangle instance. Usually reserved for lines of business, or segments.
columns
The second axis of a Triangle.
origin
The third axis of a Triangle that represents origin dates of a Triangle.
development
The fourth axis of a Triangle that represents either development age or valuation dates of a Triangle.
valuation
An implicit axis representing the valuation period of each of the cells of a
Triangle.