| Title: | Make Static HTML Website for Predictive Models |
|---|---|
| Description: | Website generator with HTML summaries for predictive models. This package uses 'DALEX' explainers to describe global model behavior. We can see how well models behave (tabs: Model Performance, Auditor), how much each variable contributes to predictions (tabs: Variable Response) and which variables are the most important for a given model (tabs: Variable Importance). We can also compare Concept Drift for pairs of models (tabs: Drifter). Additionally, data available on the website can be easily recreated in current R session. Work on this package was financially supported by the NCN Opus grant 2017/27/B/ST6/01307 at Warsaw University of Technology, Faculty of Mathematics and Information Science. |
| Authors: | Przemyslaw Biecek [aut], Magda Tatarynowicz [aut], Kamil Romaszko [aut, cre], Mateusz Urbanski [aut] |
| Maintainer: | Kamil Romaszko <[email protected]> |
| License: | Apache License 2.0 |
| Version: | 1.1 |
| Built: | 2024-11-01 03:25:22 UTC |
| Source: | https://github.com/modeloriented/modeldown |
Generates a website with HTML summaries for predictive models
modelDown( ..., modules = c("auditor", "drifter", "model_performance", "variable_importance", "variable_response"), output_folder = "output", repository_name = "repository", should_open_website = interactive() )modelDown( ..., modules = c("auditor", "drifter", "model_performance", "variable_importance", "variable_response"), output_folder = "output", repository_name = "repository", should_open_website = interactive() )
... |
one or more explainers created with |
modules |
modules that should be included in the website |
output_folder |
folder where the website will be saved |
repository_name |
name of local archivist repository that will be created |
should_open_website |
should generated website be automatically opened in default browser |
Additional arguments that could by passed by name:
remote_repository_path Path to remote repository that stores folder with archivist repository. If not provided, links to local repository will be shown.
device Device to use. Tested for "png" and "svg", but values from ggplot2::ggsave function should be working fine. Defaults to "png".
vr.vars variables which will be examined in Variable Response module. Defaults to all variables. Example vr.vars = c("var1", "var2")
vr.type types of examinations which will be conducteed in Variable Response module. Defaults to "partial". Example vr.type = c("partial", "conditional", "accumulated")
Przemysław Biecek, Magda Tatarynowicz, Kamil Romaszko, Mateusz Urbański
if(FALSE){ require("ranger") require("breakDown") require("DALEX") # Generate simple modelDown page HR_data_selected <- HR_data[1000:3000,] HR_glm_model <- glm(left~., HR_data_selected, family = "binomial") explainer_glm <- explain(HR_glm_model, data=HR_data_selected, y = HR_data_selected$left) modelDown::modelDown(explainer_glm, modules = c("model_performance", "variable_importance", "variable_response"), output_folder = tempdir(), repository_name = "HR", device = "png", vr.vars= c("average_montly_hours"), vr.type = "partial") # More complex example with all modules HR_ranger_model <- ranger(as.factor(left) ~ ., data = HR_data, num.trees = 500, classification = TRUE, probability = TRUE) explainer_ranger <- explain(HR_ranger_model, data = HR_data, y = HR_data$left, function(model, data) { return(predict(model, data)$prediction[,2]) }, na.rm=TRUE) # Two glm models used for drift detection HR_data1 <- HR_data[1:4000,] HR_data2 <- HR_data[4000:nrow(HR_data),] HR_glm_model1 <- glm(left~., HR_data1, family = "binomial") HR_glm_model2 <- glm(left~., HR_data2, family = "binomial") explainer_glm1 <- explain(HR_glm_model1, data=HR_data1, y = HR_data1$left) explainer_glm2 <- explain(HR_glm_model2, data=HR_data2, y = HR_data2$left) modelDown::modelDown(list(explainer_glm1, explainer_glm2), modules = c("auditor", "drifter", "model_performance", "variable_importance", "variable_response"), output_folder = tempdir(), repository_name = "HR", remote_repository_path = "some_user/remote_repo_name", device = "png", vr.vars= c("average_montly_hours", "time_spend_company"), vr.type = "partial") }if(FALSE){ require("ranger") require("breakDown") require("DALEX") # Generate simple modelDown page HR_data_selected <- HR_data[1000:3000,] HR_glm_model <- glm(left~., HR_data_selected, family = "binomial") explainer_glm <- explain(HR_glm_model, data=HR_data_selected, y = HR_data_selected$left) modelDown::modelDown(explainer_glm, modules = c("model_performance", "variable_importance", "variable_response"), output_folder = tempdir(), repository_name = "HR", device = "png", vr.vars= c("average_montly_hours"), vr.type = "partial") # More complex example with all modules HR_ranger_model <- ranger(as.factor(left) ~ ., data = HR_data, num.trees = 500, classification = TRUE, probability = TRUE) explainer_ranger <- explain(HR_ranger_model, data = HR_data, y = HR_data$left, function(model, data) { return(predict(model, data)$prediction[,2]) }, na.rm=TRUE) # Two glm models used for drift detection HR_data1 <- HR_data[1:4000,] HR_data2 <- HR_data[4000:nrow(HR_data),] HR_glm_model1 <- glm(left~., HR_data1, family = "binomial") HR_glm_model2 <- glm(left~., HR_data2, family = "binomial") explainer_glm1 <- explain(HR_glm_model1, data=HR_data1, y = HR_data1$left) explainer_glm2 <- explain(HR_glm_model2, data=HR_data2, y = HR_data2$left) modelDown::modelDown(list(explainer_glm1, explainer_glm2), modules = c("auditor", "drifter", "model_performance", "variable_importance", "variable_response"), output_folder = tempdir(), repository_name = "HR", remote_repository_path = "some_user/remote_repo_name", device = "png", vr.vars= c("average_montly_hours", "time_spend_company"), vr.type = "partial") }