Greedy binary splits minimising the within-node sum of squares (ESL eqs
9.10-9.13), midpoint thresholds, stopped by depth and minimum leaf size (not
pruned); matches rpart's anova splits with cp = 0.
Usage
morie_esl_decision_tree(
X,
y,
max_depth = 3,
min_leaf = 1,
min_impurity_decrease = 0
)
Arguments
- X
Predictors, n by p.
- y
Response.
- max_depth, min_leaf, min_impurity_decrease
Stopping rules.
Value
Named list: estimate (training RSS), tree (nested nodes with
1-based feature), n_leaves, depth, fitted, n, p.
References
Breiman, L., Friedman, J., Olshen, R. & Stone, C. (1984).
Classification and Regression Trees.
Examples
i <- 1:60
X <- cbind(((7 * i) %% 59) / 59 * 3, ((11 * i) %% 61) / 61 * 2)
morie_esl_decision_tree(X, sin(2 * X[, 1]) + (X[, 2] > 1.1), max_depth = 2)$n_leaves
#> [1] 4
