Rpart R决策树分数[重复]

Question

在使用 SkLearn 的 Python 中，您可以使用以下方法在决策树上创建和接收分数：

tr = tree.DecisionTreeClassifier(random_state=rseed, min_samples_split=2, ccp_alpha=0.005)
model_tree = tr.fit(train_features, train_outputs)

print(f'Model Train Accuracy: {model_tree.score(train_features, train_outputs)}')
print(f'Model Test Accuracy: {model_tree.score(test_features, test_outputs)}')

以上产生

Model Train Accuracy: 0.5942
Model Test Accuracy: 0.4933

如何使用 R 的 Rpart 在 R 中（在训练和测试数据上）获得相似的分数？

Answer 1

简而言之：

1. 如下图计算错误率
2. 确保在python和R中使用相同的参数和控制参数（见https://www.rdocumentation.org/packages/rpart/versions/4.1-15/topics/rpart.control）

model_tree <- rpart(Response ~ Predictor1 + PredictorX,
                    data = train, method = "class",
                    control = list(cp = 0.005, minsplit = 2, ...))

pred_train <- predict(model_tree, type = "class")
pred_test <- predict(model_tree, newdata = test, type = "class")

# error rate / accuracy (train set)
mean(pred_train != train$Response)

# error rate / accuracy (test set)
mean(pred_test != test$Response)