def __split(self):
"""split current node"""
max_gain, attr_index, partition_value = self.__information_gain()
self.leaf = False
self.attr_index = attr_index
self.partition_value = partition_value
tn_le = DTreeGain(DTreeSample(), self.param)
tn_le.sample.m = self.sample.m
tn_gt = DTreeGain(DTreeSample(), self.param)
tn_gt.sample.m = self.sample.m
for i in range(0, self.sample.n):
x = self.sample.X[i]
y = self.sample.Y[i]
attr_value = x[attr_index]
if attr_value <= partition_value:
tn_le.sample.add_xy(x, y)
else:
tn_gt.sample.add_xy(x, y)
left_n = len(tn_le.sample)
right_n = len(tn_gt.sample)
print >> sys.stderr, '[split]level=%d, max_gain=%f, attr_index=%d, partition_value=%f, left/right=%d/%d' % (
self.level, max_gain, attr_index, partition_value, left_n, right_n)
if left_n == 0:
tn_le = None
if right_n == 0:
tn_gt = None
return tn_le, tn_gt
def __split(self):
"""split current node"""
max_gain, attr_index, partition_value, y_value_left, y_value_right = self.__gain()
self.leaf = False
self.attr_index = attr_index
self.partition_value = partition_value
self.value = None
tn_le = DTreeLoss(DTreeSample(), self.param)
tn_le.sample.m = self.sample.m
tn_le.level = self.level + 1
tn_le.value = y_value_left
tn_gt = DTreeLoss(DTreeSample(), self.param)
tn_gt.sample.m = self.sample.m
tn_gt.level = self.level + 1
tn_gt.value = y_value_right
for i in range(0, self.sample.n):
x = self.sample.X[i]
y = self.sample.Y[i]
y_residual = self.sample.Y_residual[i]
attr_value = x[attr_index]
if attr_value <= partition_value:
tn_le.sample.add_xyr(x, y, y_residual)
else:
tn_gt.sample.add_xyr(x, y, y_residual)
tn_le.__residual_2_response()
tn_gt.__residual_2_response()
left_n = len(tn_le.sample)
right_n = len(tn_gt.sample)
print >>sys.stderr, '[split]level=%d, max_gain=%f, attr_index=%d, partition_value=%f, left/right=%d/%d' % (self.level, max_gain, attr_index, partition_value, left_n, right_n)
if left_n == 0:
tn_le = None
if right_n == 0:
tn_gt = None
return tn_le, tn_gt
# -*- coding: utf-8 -*- # # 预测房地产价格 # # author: yafei([email protected]) # import sys import codecs import locale from dtree_loss import DTreeLoss from dtree_parameter import DTreeParameter from dtree_sample import DTreeSample if __name__ == '__main__': param = DTreeParameter() sample = DTreeSample() sample.load('real-estate.txt') dt = DTreeLoss(sample, param) dt.train(None) feature_map = { 0: u'结构', 1: u'装修', 2: u'周边', 3: u'地段', 4: u'绿化', 5: u'交通', 6: u'户均车位', } # 为了输出中文 locale.setlocale(locale.LC_ALL, '')
