8.46 Build a regression tree
We can reuse class_tree_spec as a base for the regression decision tree specification.
reg_tree_spec <- class_tree_spec %>%
set_mode("regression")We are using the Carseats dataset. Let’s do the validation split.
set.seed(1010)
carseats_split <- initial_split(Carseats)
carseats_train <- training(carseats_split)
carseats_test <- testing(carseats_split)Fit the decision tree regression model
reg_tree_fit <- fit(reg_tree_spec, Sales ~ ., data = carseats_train)
reg_tree_fit## parsnip model object
##
## n= 300
##
## node), split, n, deviance, yval
## * denotes terminal node
##
## 1) root 300 2529.534000 7.469333
## 2) ShelveLoc=Bad,Medium 235 1476.666000 6.695702
## 4) Price>=124.5 86 393.091600 5.283721
## 8) CompPrice< 147.5 72 303.087600 4.916528
## 16) Price>=137.5 27 99.462070 3.735556 *
## 17) Price< 137.5 45 143.374700 5.625111
## 34) Advertising< 14.5 38 99.190880 5.249211 *
## 35) Advertising>=14.5 7 9.665971 7.665714 *
## 9) CompPrice>=147.5 14 30.370240 7.172143 *
## 5) Price< 124.5 149 813.155300 7.510671
## 10) Age>=50.5 91 411.524700 6.751758
## 20) Price>=86.5 80 283.338300 6.353750
## 40) CompPrice< 123.5 52 157.669400 5.689231
## 80) Price>=102.5 24 52.750780 4.710833
## 160) ShelveLoc=Bad 8 12.693000 3.120000 *
## 161) ShelveLoc=Medium 16 9.688775 5.506250 *
## 81) Price< 102.5 28 62.252070 6.527857 *
## 41) CompPrice>=123.5 28 60.061870 7.587857 *
## 21) Price< 86.5 11 23.347450 9.646364 *
## 11) Age< 50.5 58 266.987700 8.701379
## 22) Income< 59.5 18 65.389180 7.101111 *
## 23) Income>=59.5 40 134.760100 9.421500 *
## 3) ShelveLoc=Good 65 403.719500 10.266310
## 6) Price>=109.5 42 197.649800 9.155238
## 12) Price>=142.5 12 36.647220 7.152500 *
## 13) Price< 142.5 30 93.618500 9.956333
## 26) Age>=61.5 9 17.323960 8.537778 *
## 27) Age< 61.5 21 50.422110 10.564290
## 54) CompPrice< 129.5 11 14.599670 9.515455 *
## 55) CompPrice>=129.5 10 10.411360 11.718000 *
## 7) Price< 109.5 23 59.542770 12.295220 *