model_compare: Compare two or more insuRglm models

In realgabon/insuRglm: Tools for GLM Modeling in Insurance Context

Compare two or more insuRglm models

Description

Compares multiple insuRglm models with the current (last) model. The comparison models must be saved using model_save.

Usage

model_compare(setup, type = c("rmse", "nested_model_test"), buckets = 10)

Arguments

setup	Setup object. Created at the start of the workflow. Usually piped in from previous step.
type	Character scalar. One of 'nested_model_test' or 'rmse'. 'nested_model_test' will produce nested model test in form of matrix of standard error percentages. 'rmse' will produce a plot comparing RMSE across all models (also separately for train and CV if model_crossval was used)

Value

Either a matrix-like dataframe of nested model test results or a ggplot2 chart.

See Also

model_save, model_crossval

Examples

require(dplyr) # for the pipe operator
data('sev_train')

setup <- setup(
  data_train = sev_train,
  target = 'sev',
  weight = 'numclaims',
  family = 'gamma',
  keep_cols = c('pol_nbr', 'exposure', 'premium')
)

modeling <- setup %>%
  factor_add(pol_yr) %>%
  model_fit() %>%
  model_save('model1') %>%
  factor_add(agecat) %>%
  model_fit() %>%
  model_save('model2') %>%
  factor_modify(agecat = variate(agecat, type = 'non_prop', mapping = c(1, 2, 3, 4, 5, 6))) %>%
  model_fit()

# nested model test of model with and without the agecat
modeling %>%
  model_compare(type = 'nested_model_test')

# compare training RMSE on all models so far
modeling %>%
  model_compare(type = 'rmse')

# compare both training and CV RMSE on all models
modeling_with_cv <- modeling %>%
  model_crossval(cv_folds = 10, stratified = TRUE)

modeling_with_cv %>%
  model_compare(type = 'rmse')

realgabon/insuRglm documentation built on Jan. 2, 2023, 2:51 a.m.