def __init__(self, **kwargs):
super(RecSys2ClickSession, self).__init__(X=None, y=None, **kwargs)
# df = pandas.read_csv('/home/wuzz/ihpc/dataset/yoochoose_data/2_clicks_sess.csv')
FEATURE = [
'Price', 'ItemMaxPrice', 'ItemMinPrice', 'ItemTotalClicks',
'ItemTotalBuys', 'ItemBuyingProbability', 'cat1', 'cat2', 'cat3',
'cat4', 'cat5', 'cat6', 'cat7', 'cat8', 'cat9', 'cat10', 'cat11',
'cat12', 'ItemDuration', 'ItemMonth', 'ItemDay', 'ItemHourInMins',
'Sales'
]
LABEL = ['BuyInSession']
df = pandas.read_csv(
'/home/wuzz/ihpc/dataset/yoochoose_data/2_clicks_sess.csv')
df = df.sort('SessionID')[:1000000]
data = Dataset_by_Session(df=df,
sort_by_session=True,
train_valid_test=self.ratio)
train_df, valid_df, test_df = data.split_df()
# import pdb; pdb.set_trace()
self.set_train(X=train_df[FEATURE].values,
y=make_one_hot(
train_df[LABEL].values.reshape(len(train_df)), 2))
self.set_valid(X=valid_df[FEATURE].values,
y=make_one_hot(
valid_df[LABEL].values.reshape(len(valid_df)), 2))
self.set_test(X=test_df[FEATURE].values,
y=make_one_hot(
test_df[LABEL].values.reshape(len(test_df)), 2))
def __init__(self, **kwargs):
super(RecSys, self).__init__(X=None, y=None, **kwargs)
csv_path = '/home/wuzz/ihpc/onlinebehavior/sessions.csv'
data = Dataset_by_Session(csv_path=csv_path, train_valid_test = self.ratio)
tbl = data.split(features=data.feature, labels=data.label)
self.set_train(X=tbl['train_fea'], y=make_one_hot(tbl['train_lbl'], 2))
self.set_valid(X=tbl['valid_fea'], y=make_one_hot(tbl['valid_lbl'], 2))
self.set_test(X=tbl['test_fea'], y=make_one_hot(tbl['test_lbl'], 2))
def __init__(self, **kwargs):
super(RecSysJitter, self).__init__(X=None, y=None, **kwargs)
data_dir = '/home/wuzz/recsys2015/data/jitter_stdr0_1_dup_15'
train = Dataset_by_Session(sort_by_session=False, csv_path=data_dir + '/train_jitter.csv')
valid = Dataset_by_Session(sort_by_session=False, csv_path=data_dir + '/valid.csv')
test = Dataset_by_Session(sort_by_session=False, csv_path=data_dir + '/test.csv')
self.set_train(X=train.feature, y=make_one_hot(train.label,2))
self.set_valid(X=valid.feature, y=make_one_hot(valid.label,2))
self.set_test(X=test.feature, y=make_one_hot(test.label,2))
def __init__(self, **kwargs):
super(RecSys, self).__init__(X=None, y=None, **kwargs)
csv_path = '/home/wuzz/ihpc/onlinebehavior/sessions.csv'
data = Dataset_by_Session(csv_path=csv_path,
train_valid_test=self.ratio)
tbl = data.split(features=data.feature, labels=data.label)
self.set_train(X=tbl['train_fea'], y=make_one_hot(tbl['train_lbl'], 2))
self.set_valid(X=tbl['valid_fea'], y=make_one_hot(tbl['valid_lbl'], 2))
self.set_test(X=tbl['test_fea'], y=make_one_hot(tbl['test_lbl'], 2))
def __init__(self, **kwargs):
super(RecSysJitter, self).__init__(X=None, y=None, **kwargs)
data_dir = '/home/wuzz/recsys2015/data/jitter_stdr0_1_dup_15'
train = Dataset_by_Session(sort_by_session=False,
csv_path=data_dir + '/train_jitter.csv')
valid = Dataset_by_Session(sort_by_session=False,
csv_path=data_dir + '/valid.csv')
test = Dataset_by_Session(sort_by_session=False,
csv_path=data_dir + '/test.csv')
self.set_train(X=train.feature, y=make_one_hot(train.label, 2))
self.set_valid(X=valid.feature, y=make_one_hot(valid.label, 2))
self.set_test(X=test.feature, y=make_one_hot(test.label, 2))
def save_gaussian_one_hot(y, save_dir, std):
print 'saving'
one_hot_y = make_one_hot(y, 1998)
gaussian = Gaussian(std=std)
one_hot_y = gaussian.apply(one_hot_y)
with open("%s/sample_y_onehot_gaussian_noise_std%s.npy"%(save_dir,str(std)), 'wb') as yout:
np.save(yout, one_hot_y)
print 'saving done!'
def __init__(self, **kwargs):
super(RecSysPosterior, self).__init__(X=None, y=None, **kwargs)
sav_dir = '/home/wuzz/ihpc/dataset/posteriors'
with open(sav_dir + '/train_y.npy') as train_y_fin:
train_X = np.load(train_y_fin)
with open(sav_dir + '/train_lbl.npy') as train_lbl_fin:
train_y = np.load(train_lbl_fin)
with open(sav_dir + '/test_y.npy') as test_y_fin:
test_X = np.load(test_y_fin)
with open(sav_dir + '/test_lbl.npy') as test_lbl_fin:
test_y = np.load(test_lbl_fin)
self.set_train(X=train_X, y=make_one_hot(train_y, 2))
self.set_valid(X=test_X, y=make_one_hot(test_y, 2))
self.set_test(X=test_X, y=make_one_hot(test_y, 2))
def __init__(self, **kwargs):
super(RecSysPosterior, self).__init__(X=None, y=None, **kwargs)
sav_dir = '/home/wuzz/ihpc/dataset/posteriors'
with open(sav_dir + '/train_y.npy') as train_y_fin:
train_X = np.load(train_y_fin)
with open(sav_dir + '/train_lbl.npy') as train_lbl_fin:
train_y = np.load(train_lbl_fin)
with open(sav_dir + '/test_y.npy') as test_y_fin:
test_X = np.load(test_y_fin)
with open(sav_dir + '/test_lbl.npy') as test_lbl_fin:
test_y = np.load(test_lbl_fin)
self.set_train(X=train_X, y=make_one_hot(train_y, 2))
self.set_valid(X=test_X, y=make_one_hot(test_y, 2))
self.set_test(X=test_X, y=make_one_hot(test_y, 2))
def __init__(self, **kwargs):
dir = '/home/stuwzhz/datasets/spectral-features/npy2'
with open('%s/sample_y.npy'%dir) as yin, \
open('%s/sample_X.npy'%dir) as Xin:
y = np.load(yin)
y = make_one_hot(y, 1998)
X = np.load(Xin)
super(I2R_Posterior_NoisyFeat_Sample, self).__init__(X=X, y=y, **kwargs)
def save_gaussian_one_hot(y, save_dir, std):
print 'saving'
one_hot_y = make_one_hot(y, 1998)
gaussian = Gaussian(std=std)
one_hot_y = gaussian.apply(one_hot_y)
with open(
"%s/sample_y_onehot_gaussian_noise_std%s.npy" %
(save_dir, str(std)), 'wb') as yout:
np.save(yout, one_hot_y)
print 'saving done!'
def __init__(self, **kwargs):
super(RecSys2ClickSession, self).__init__(X=None, y=None, **kwargs)
# df = pandas.read_csv('/home/wuzz/ihpc/dataset/yoochoose_data/2_clicks_sess.csv')
FEATURE = ['Price', 'ItemMaxPrice', 'ItemMinPrice', 'ItemTotalClicks', 'ItemTotalBuys', 'ItemBuyingProbability',
'cat1', 'cat2', 'cat3', 'cat4', 'cat5', 'cat6', 'cat7', 'cat8', 'cat9', 'cat10', 'cat11', 'cat12',
'ItemDuration', 'ItemMonth', 'ItemDay', 'ItemHourInMins', 'Sales']
LABEL = ['BuyInSession']
df = pandas.read_csv('/home/wuzz/ihpc/dataset/yoochoose_data/2_clicks_sess.csv')
df = df.sort('SessionID')[:1000000]
data = Dataset_by_Session(df=df,
sort_by_session=True,
train_valid_test=self.ratio)
train_df, valid_df, test_df = data.split_df()
# import pdb; pdb.set_trace()
self.set_train(X=train_df[FEATURE].values, y=make_one_hot(train_df[LABEL].values.reshape(len(train_df)), 2))
self.set_valid(X=valid_df[FEATURE].values, y=make_one_hot(valid_df[LABEL].values.reshape(len(valid_df)), 2))
self.set_test(X=test_df[FEATURE].values, y=make_one_hot(test_df[LABEL].values.reshape(len(test_df)), 2))
def __init__(self, one_hot=False, **kwargs):
dir = '/Volumes/Storage/Unilever_Challenge/dataset'
with open(dir + '/train.npy') as Xin:
data = np.load(Xin)
X, y = self.make_Xy(data)
if one_hot:
y = make_one_hot(y, 8)
else:
y = y.reshape((y.shape[0], 1))
super(Unilever, self).__init__(X=X, y=y, **kwargs)
def __init__(self, **kwargs):
dir = '/home/stuwzhz/datasets/spectral-features/npy2'
with open('%s/sample_y.npy'%dir) as yin, \
open('%s/sample_X.npy'%dir) as Xin:
y = np.load(yin)
y = make_one_hot(y, 1998)
X = np.load(Xin)
super(I2R_Posterior_NoisyFeat_Sample, self).__init__(X=X,
y=y,
**kwargs)
def __init__(self, one_hot=False, **kwargs):
dir = '/Volumes/Storage/Unilever_Challenge/dataset'
with open(dir + '/train.npy') as Xin:
data = np.load(Xin)
X, y = self.make_Xy(data)
if one_hot:
y = make_one_hot(y, 8)
else:
y = y.reshape((y.shape[0], 1))
super(Unilever, self).__init__(X=X, y=y, **kwargs)
def __init__(self, **kwargs):
dir = '/home/stuwzhz/datasets/spectral-features/npy2'
with open('%s/sample_y.npy'%dir) as yin:
y = np.load(yin)
y_tmp = []
for e in y:
if e > 150:
y_tmp.append(e)
y_tmp = np.asarray(y_tmp)
y_tmp = make_one_hot(y_tmp, 1998)
super(I2R_Posterior_Gaussian_Noisy_Sample, self).__init__(X=y_tmp, y=y_tmp, **kwargs)
def __init__(self, **kwargs):
dir = '/home/stuwzhz/datasets/spectral-features/npy2'
with open('%s/sample_y.npy' % dir) as yin:
y = np.load(yin)
y_tmp = []
for e in y:
if e > 150:
y_tmp.append(e)
y_tmp = np.asarray(y_tmp)
y_tmp = make_one_hot(y_tmp, 1998)
super(I2R_Posterior_Gaussian_Noisy_Sample, self).__init__(X=y_tmp,
y=y_tmp,
**kwargs)
