def test_model():
data_file = './data/feature_data.csv'
dateset = DataSet(data_file)
gbdt = GBDT.load_model("./", "test")
loss = 0.0
err_num = 0
for item_id in dateset.get_instances_idset():
predict = gbdt.predict(dateset.instances[item_id])
print "predict", predict, dateset.get_instance(item_id)['label']
y_i = dateset.get_instance(item_id)['label']
f_value = predict
p_1 = 1 / (1 + exp(-2 * f_value))
loss -= ((1 + y_i) * log(p_1) / 2) + ((1 - y_i) * log(1 - p_1) / 2)
if abs(y_i - predict) > 0.5:
err_num += 1
print "loss %s" % (loss / dateset.size())
print "err_classifier rate: %s" % ((err_num * 1.0) / dateset.size())
"""
如下为credit.data.csv文件的训练信息
iter1 : train loss=0.371342
iter2 : train loss=0.238326
iter3 : train loss=0.163624
iter4 : train loss=0.123063
iter5 : train loss=0.087872
iter6 : train loss=0.065684
iter7 : train loss=0.049936
iter8 : train loss=0.041866
iter9 : train loss=0.035695
iter10 : train loss=0.030581
iter11 : train loss=0.027034
iter12 : train loss=0.024570
iter13 : train loss=0.019227
iter14 : train loss=0.015794
iter15 : train loss=0.013484
iter16 : train loss=0.010941
iter17 : train loss=0.009879
iter18 : train loss=0.008619
iter19 : train loss=0.007306
iter20 : train loss=0.005610
"""
from gbdt.data import DataSet
from gbdt.model import GBDT
if __name__ == '__main__':
data_file = './data/credit.data.csv'
dateset = DataSet(data_file)
gbdt = GBDT(max_iter=20, sample_rate=0.8, learn_rate=0.5, max_depth=7, loss_type='binary-classification')
gbdt.fit(dateset, dateset.get_instances_idset())
def train_model():
data_file = './data/feature_data.csv'
dateset = DataSet(data_file)
gbdt = GBDT(max_iter=80,
sample_rate=0.8,
learn_rate=0.1,
max_depth=7,
loss_type='regression')
gbdt.fit(dateset, set(list(dateset.get_instances_idset())[:1200]))
GBDT.save_model(gbdt, "./", "test")
predict = gbdt.predict(dateset.instances[1])
print "predict", predict, dateset.get_instance(1)['label']
print "#########################"
predict = gbdt.predict(dateset.instances[2])
print "predict", predict, dateset.get_instance(2)['label']
print "#########################"
predict = gbdt.predict(dateset.instances[3])
print "predict", predict, dateset.get_instance(3)['label']
print "#########################"
predict = gbdt.predict(dateset.instances[4])
print "predict", predict, dateset.get_instance(4)['label']
predict = gbdt.predict(dateset.instances[402])
print "predict", predict, dateset.get_instance(402)['label']
iter3 : train loss=0.163624
iter4 : train loss=0.123063
iter5 : train loss=0.087872
iter6 : train loss=0.065684
iter7 : train loss=0.049936
iter8 : train loss=0.041866
iter9 : train loss=0.035695
iter10 : train loss=0.030581
iter11 : train loss=0.027034
iter12 : train loss=0.024570
iter13 : train loss=0.019227
iter14 : train loss=0.015794
iter15 : train loss=0.013484
iter16 : train loss=0.010941
iter17 : train loss=0.009879
iter18 : train loss=0.008619
iter19 : train loss=0.007306
iter20 : train loss=0.005610
"""
from gbdt.data import DataSet
from gbdt.model import GBDT
if __name__ == '__main__':
data_file = './data/credit.data.csv'
dateset = DataSet(data_file)
gbdt = GBDT(max_iter=20,
sample_rate=0.8,
learn_rate=0.5,
max_depth=7,
loss_type='binary-classification')
gbdt.fit(dateset, dateset.get_instances_idset())
from gbdt.model import GBDT
from gbdt.data import DataSet
model = GBDT(tree_depth=3, learning_rate=0.01, max_iter=2000)
dataset = DataSet('data/ages.csv', 'Age')
model.fit(dataset)
x = {'LikesGardening': False, 'PlaysVideoGames': True, 'LikesHats': False}
print model.predict(x)