Package index

Feature attribution methods

All implemented feature attribution methods for the survival neural networks ‘DeepHit’, ‘DeepSurv’, and ‘CoxTime’.

surv_grad()

Calculate the Gradient of the Survival Function

surv_smoothgrad()

Calculate the SmoothGrad values of the Survival Function

surv_intgrad()

Calculate the Integrated Gradients of the Survival Function

surv_gradSHAP()

Calculate the GradSHAP values of the Survival Function

Creating explainer object

The basic building block of the package is the ‘explainer’ object which contains the model and the data needed for the feature attribution methods.

explain()

Explain a model

Extracting a model from survivalmodels

The ‘extract_model’ function extracts the model from the ‘survivalmodels’ object and returns all necessary information for the survival model to be explained.

extract_model()

Extract model information from a survivalmodels object

Plotting and printing

The ‘plot’ function is used to plot the feature attribution results in different ways. The ‘plot’ function is a generic function that calls the appropriate method depending on the argument ‘type’. The ‘print’ function is used to print extracted model, explainers, and feature attribution results..

plot(<surv_result>) plot_force() plot_pred() plot_attr() plot_contr()

Plot Methods for Survival Attribution Results

print(<surv_result>)

Print function for surv_result objects

print(<explainer_coxtime>) print(<explainer_deepsurv>) print(<explainer_deephit>)

Custom print method for explainer objects

print(<extracted_survivalmodels_coxtime>) print(<extracted_survivalmodels_deepsurv>) print(<extracted_survivalmodels_deephit>)

Print method for extracted pycox survival model

Data transformation

These functions are used to transform the data into a data.frame or a data.table format.

as.data.frame(<surv_result>) as.data.table.surv_result()

Convert survival attribution results to a data.frame