def get_confs1():
score = 5
return [
MyConfStableLTa(classifier=cnn(nb_epoch=5), score=score),
MyConfStableLTa(classifier=cnn(),score=score),
MyConfStableLTa(classifier=cnn(num_filt_2=4), score=score),
MyConfStableLTa(classifier=cnn(num_filt_1=6, num_filt_2=4), score=score),
MyConfStableLTa(classifier=cnn(batch_size=1000), score=score),
]
def get_confs():
score = 5
return [
#MyConfStableLTa(classifier=ccl2(batch_size=32, nb_epoch=20), score=score),
#MyConfStableLTa(classifier=cnn(batch_size=32, nb_epoch=20), score=score),
MyConfStableLTa(classifier=Logit2(), score=score),
]
import os
import pandas as pd
local_path = os.path.dirname(__file__)
root = os.path.join(local_path, '..', "..")
sys.path.append(root)
from main.work.conf import MyConfStableLTa
def work(confer):
if not os.path.exists(confer.get_sel_file()) or confer.force:
df = confer.selector.work()
print(df.shape)
df.to_pickle(confer.get_sel_file())
if __name__ == "__main__":
from main.classifier.tree import MySGDClassifier
confer = MyConfStableLTa(classifier=MySGDClassifier(),score=5)
ta_file = pd.read_pickle(confer.get_bitlize_file())
ta_file1 = confer.selector._select(ta_file,
confer.model_split.train_start,
confer.model_split.train_end,
confer.scores[0].get_name())
ta_file2 = confer.selector._select(ta_file,
"2013-01-01",
"2014-01-01",
confer.scores[0].get_name())
ta = ta_file2.merge(ta_file1, left_index=True, right_index=True)
print(ta.head())
#"adj": MyConfStableLTa(classifier=Logit2(30), is_adj= True),
#"score5_30": MyConfStableLTa(classifier=Logit2(dim=30), is_adj = False),
#"score5_64": MyConfStableLTa(classifier=Logit2(dim=64), is_adj = False),
#"score5_8": MyConfStableLTa(classifier=Logit2(dim=8), is_adj = False),
#"score5_16": MyConfStableLTa(classifier=Logit2(dim=16), is_adj = False),
#"score5_24": MyConfStableLTa(classifier=Logit2(dim=24), is_adj = False),
#"score5_32": MyConfStableLTa(classifier=Logit2(dim=32), is_adj = False),
#"score5_40": MyConfStableLTa(classifier=Logit2(dim=40), is_adj = False),
#"score5_40_2": MyConfStableLTa(classifier=Logit2(dim=40, dropout=0.2), is_adj = False),
#"score5_40_4": MyConfStableLTa(classifier=Logit2(dim=40, dropout=0.4), is_adj = False),
#"score5_40_6": MyConfStableLTa(classifier=Logit2(dim=40, dropout=0.6), is_adj = False),
#"score5_40_6_1": MyConfStableLTa(classifier=Logit2(hs=1, dim=40, dropout=0.6), is_adj = False),
#"score5_40_6_2": MyConfStableLTa(classifier=Logit2(hs=2, dim=40, dropout=0.6), is_adj = False),
#"score5_40_6_3": MyConfStableLTa(classifier=Logit2(hs=3, dim=40, dropout=0.6), is_adj = False), # best
"score5_40_6_3_adam1":
MyConfStableLTa(classifier=Logit3(hs=3, dim=40, dropout=0.6, lr=4e-5),
is_adj=False), # best
"score5_40_6_3_adam2":
MyConfStableLTa(classifier=Logit3(hs=3, dim=40, dropout=0.6, lr=8e-5),
is_adj=False), # best
#"score5_40_6_3_adam3": MyConfStableLTa(classifier=Logit3(hs=3, dim=40, dropout=0.6, lr=2e-5), is_adj = False), # best
#"score5_40_6_3_adam4": MyConfStableLTa(classifier=Logit3(hs=3, dim=40, dropout=0.6, lr=1e-6), is_adj = False), # best
#"score5_40_6_4": MyConfStableLTa(classifier=Logit2(hs=4, dim=40, dropout=0.6), is_adj = False),
#"score5_40_6_5": MyConfStableLTa(classifier=Logit2(hs=5, dim=40, dropout=0.6), is_adj = False),
#"score5_40_6_6": MyConfStableLTa(classifier=Logit2(hs=6, dim=40, dropout=0.6), is_adj = False),
#"score5_40_8": MyConfStableLTa(classifier=Logit2(dim=40, dropout=0.8), is_adj = False),
#"score5_40_10": MyConfStableLTa(classifier=Logit2(dim=40, dropout=1.0), is_adj = False),
#"score5_48": MyConfStableLTa(classifier=Logit2(dim=48), is_adj = False),
#"ccl2": MyConfStableLTa(classifier=Logit2(), is_adj = False),
#"ccl3": MyConfStableLTa(classifier=Logit2(dim=10, hs=6), is_adj = False),
#"delta": MyConfStableLTa(classifier=Logit(nb_epoch=10), is_adj = False),
#"delta2": MyConfStableLTa(classifier=Logit(nb_epoch=20), is_adj = False),
import pickle
import pandas as pd
import numpy as np
local_path = os.path.dirname(__file__)
root = os.path.join(local_path, '..',"..")
sys.path.append(root)
from main import base
from main.base.score2 import ScoreLabel
from main.work import model
from main.work import build
from main.work.conf import MyConfStableLTa
from main.model import ana
confer = MyConfStableLTa()
#build.work(confer)
#model.work(confer)
df = pd.read_pickle(os.path.join(root, 'output', "result_20170205.pkl"))
print(ana.roc_auc(df, confer))
clazz_file_name = confer.get_classifier_file()
with open(clazz_file_name, 'rb') as fin:
clazz = pickle.load(fin)
feat_names = base.get_feat_names(df)
ipts = sorted(clazz.get_feature_importances(feat_names).items(), key=lambda a:a[1], reverse=True)
def get_confs2():
score = 5
return [
MyConfStableLTa(classifier=ccl2(batch_size=32, nb_epoch=10),
score=score),
]
def get_confs2():
score = 5
return [
MyConfStableLTa(classifier=Ts(max_iterations=20000), score=score),
]
import pickle
import pandas as pd
import numpy as np
local_path = os.path.dirname(__file__)
root = os.path.join(local_path, '..', "..")
sys.path.append(root)
from main import base
from main.base.score2 import ScoreLabel
from main.work import model
from main.work import build
from main.work.conf import MyConfStableLTa
from main.model import ana
confer = MyConfStableLTa()
#build.work(confer)
#model.work(confer)
df = pd.read_pickle(os.path.join(root, 'output', "result_20170205.pkl"))
print(ana.roc_auc(df, confer))
clazz_file_name = confer.get_classifier_file()
with open(clazz_file_name, 'rb') as fin:
clazz = pickle.load(fin)
feat_names = base.get_feat_names(df)
ipts = sorted(clazz.get_feature_importances(feat_names).items(),
key=lambda a: a[1],
reverse=True)
iter_num = 2
abtest_models = {
#"Logit10":Logit2(nb_epoch=10),
"Logit20":Logit2(nb_epoch=20),
"Logit30":Logit2(nb_epoch=30),
#"Logit30-10":Logit2(nb_epoch=30, hs=10),
#"Logit40":Logit2(nb_epoch=40),
"Logit50":Logit2(nb_epoch=50),
#"Logit80":Logit2(nb_epoch=80),
#"MDN" : MyMdnClassifier(),
}
abtest_confs = {
#"adj": MyConfStableLTa(classifier=Logit2(30), is_adj= True),
#"score5": MyConfStableLTa(classifier=Logit2(30), is_adj = False),
"score4": MyConfStableLTa(classifier=Logit2(30), is_adj = False, score=4),
"score6": MyConfStableLTa(classifier=Logit2(30), is_adj = False, score=6),
#"score2": MyConfStableLTa(classifier=Logit2(30), is_adj = False, score=2),
#"score8": MyConfStableLTa(classifier=Logit2(30), is_adj = False, score=8),
#"score32": MyConfStableLTa(classifier=Logit2(30), is_adj = False, score=32),
}
result_dict = {}
from optparse import OptionParser
parser = OptionParser()
parser.add_option('-f', '--force', action='store_true',default = False, dest='force', help = 'do not use any tmp file')
(options, args) = parser.parse_args()
report_file = os.path.join(root, 'data', 'report', 'abtest.txt')
fd = open(report_file, "w")
for model_name in abtest_confs.keys():