def __init__(self,
n_estimator=10,
learning_rate=0.01,
min_sample=2,
min_gain=0.1,
max_depth=10):
super(GBDTRegressionScratch, self).__init__(n_estimator, learning_rate)
# 回归树损失函数维平方损失
self._loss = SquareLoss()
for _ in range(self._n_estimator):
tree = CARTRegressionScratch(min_sample, min_gain, max_depth)
self._trees.append(tree)
class GBDTRegressionScratch(GBDTScratch):
def __init__(self,
n_estimator=10,
learning_rate=0.01,
min_sample=2,
min_gain=0.1,
max_depth=10):
super(GBDTRegressionScratch, self).__init__(n_estimator, learning_rate)
# 回归树损失函数维平方损失
self._loss = SquareLoss()
for _ in range(self._n_estimator):
tree = CARTRegressionScratch(min_sample, min_gain, max_depth)
self._trees.append(tree)
def fit(self, X, y):
"""模型训练"""
n_sample = y.shape[0]
residual_pred = np.zeros(n_sample)
for i in range(self._n_estimator):
residual_gradient = self._loss.calc_gradient(y, residual_pred)
# 每棵树以残差为目标进行训练
self._trees[i].fit(X, residual_gradient)
residual_update = np.zeros(n_sample)
for j in range(n_sample):
residual_update[j] = self._trees[i].predict(X[j])
residual_pred -= self._lr * residual_update
def predict(self, x):
"""给定输入样本，预测输出"""
y_pred = 0
for tree in self._trees:
residual_update = tree.predict(x)
y_pred -= self._lr * residual_update
return y_pred