What If 2D Plots

Przemyslaw Biecek

2018-12-13

Introduction

Here we present 2D What If plots.

First, an example for regression model.

library("DALEX")
library("ceterisParibus")
library("randomForest")
set.seed(59)

apartments_rf_model <- randomForest(m2.price ~ construction.year + surface + floor + no.rooms + district, data = apartments)

explainer_rf <- explain(apartments_rf_model,
      data = apartmentsTest[,2:6], y = apartmentsTest$m2.price)

new_apartment <- apartmentsTest[1, ]
new_apartment

##      m2.price construction.year surface floor no.rooms    district
## 1001     4644              1976     131     3        5 Srodmiescie

wi_rf_2d <- what_if_2d(explainer_rf, observation = new_apartment)
wi_rf_2d

##           y_hat new_x1 new_x2            vname1  vname2        label
## 1001   4899.536   1920   20.0 construction.year surface randomForest
## 1001.1 4896.920   1920   21.3 construction.year surface randomForest
## 1001.2 4900.060   1920   22.6 construction.year surface randomForest
## 1001.3 4900.187   1920   23.9 construction.year surface randomForest
## 1001.4 4901.809   1920   25.2 construction.year surface randomForest
## 1001.5 4896.952   1920   26.5 construction.year surface randomForest

plot(wi_rf_2d)

plot(wi_rf_2d, add_contour = FALSE)

plot(wi_rf_2d, add_observation = FALSE)

plot(wi_rf_2d, add_raster = FALSE)

And an example for classification.

# HR data
model <- randomForest(status ~ gender + age + hours + evaluation + salary, data = HR)
pred1 <- function(m, x)   predict(m, x, type = "prob")[,1]
explainer_rf_fired <- explain(model, data = HR[,1:5],
   y = HR$status == "fired",
   predict_function = pred1, label = "fired")

new_emp <- HR[1, ]
new_emp

##   gender      age    hours evaluation salary status
## 1   male 32.58267 41.88626          3      1  fired

wi_rf_2d <- what_if_2d(explainer_rf_fired, observation = new_emp)
wi_rf_2d

##     y_hat   new_x1   new_x2 vname1 vname2 label
## 1   0.472 20.00389 35.00000    age  hours fired
## 1.1 0.524 20.00389 35.44978    age  hours fired
## 1.2 0.582 20.00389 35.89955    age  hours fired
## 1.3 0.570 20.00389 36.34933    age  hours fired
## 1.4 0.540 20.00389 36.79911    age  hours fired
## 1.5 0.582 20.00389 37.24889    age  hours fired

plot(wi_rf_2d)