def get_result_path():
cur_dir = get_root_path()
dir_path = 'datas/' + MyCode.get() + '/'
cur_dir += '/'
cur_dir += dir_path
agl.createDir(cur_dir)
return cur_dir
def run():
fpath = env.get_root_path()
for code in stock.get_codes():
fname = '/datasources/%s.csv' % (code)
fname = fpath + fname
df = stock.getFiveHisdatDf(code)
df.to_csv(fname)
def load_user_info():
"""json [user,pwd]"""
fname = os.path.abspath(get_root_path()+'/../notify/'+'info.json')
f = open(fname, 'r')
t = str(f.read())
f.close()
t = t.replace('\t','')
t = t.replace('\n', '')
return json.loads(t)
def __init__(self, reinit=False):
self.fname = os.path.join(env.get_root_path(), 'datas', self.fname)
if os.path.exists(self.fname) and reinit == False:
self.df = pd.read_csv(self.fname)
#print(self.df)
else:
result = self._initLabeDescTable()
self.df = pd.DataFrame(result)
self.df.to_csv(self.fname)
print(self.df)
print('write label_desc_table')
def getLocal(code):
import pandas as pd
from autoxd.cnn_boll import env
fname = '../datas/%s.csv' % (code)
fname = env.get_root_path() + '/datas/%s.csv' % (code)
import os
print(os.path.abspath(fname))
df = pd.read_csv(fname)
df.index = pd.DatetimeIndex(df[df.columns[0]])
df = stock.TDX_BOLL_df(df)
return df
def genImgToFile(code):
"""产生图形到文件
"""
datas = load_data()
# indexs: list 数据偏移索引
df = pd.read_csv(get_result_csv_path(), index_col=0)
indexs = df['datas_index'].values
cur_dir = get_root_path()
fname = cur_dir + '/img_labels/imgs'
if not os.path.exists(fname):
agl.createDir(fname)
for i in indexs:
i = int(i)
fname1 =fname + '/%s_%d.png'%(code, i)
print(fname1)
pl.figure
draw(datas[i])
pl.savefig(fname1)
pl.close()
def load_data(num=-1, method='img'):
"""加载imgs
method: str img/data
return: (x_train, y_train), (x_test, y_test)
x_train, np.dnarray (num, row, col)
y_train, (num, [0,0,0,1,0,0]) 分类标签
"""
img_path = os.path.join(env.get_root_path(), 'img_labels/imgs/')
files = os.listdir(img_path)
files = files[:num]
datas = None
pre_code = ''
imgs = []
labels = []
n = 28
label_converter = Label2Id()
for f in files:
fname = img_path + f
#label
#这里和pearson_clust里的数据加载有区别， 这里是遍历
f = f.split('.')[0]
code, datas_index = str(f).split('_')
print(code)
datas_index = int(datas_index)
label_path = os.path.join(env.get_root_path(),
('datas/%s/%s') % (code, g_fname_csv))
#... 等待人工标签结果， 人工标签最后再进行归类
table_colmns = "id,datas_index,code,dt,tick_period,clust_id,label_id,label_desc".split(
',')
df = pd.read_csv(label_path)
label = df[df[table_colmns[1]] == int(datas_index)]
label_id = np.nan
if len(label) > 0:
label = label[table_colmns[-1]].values[0]
if isinstance(label, str) and label[-1] == ',':
label = label[:-1]
label_id = label_converter.label_desc_to_label_id(label)
labels.append(label_id)
#img
if method == 'img':
img = cv2.imread(fname, cv2.IMREAD_GRAYSCALE)
img = cv2.resize(img, (64 * 3, 48 * 3))
img = np.array(img)
img[img == 255] = 0
if method == 'data':
if pre_code != code:
datas = main_load_data(code)
#归一化
bolls = datas[datas_index]
img = BollsToImg(bolls)
img = img.astype(np.uint8)
#print(img)
imgs.append(img)
#for i in range(5):
#imgs += imgs
#labels += labels
data = np.array(imgs)
labels = np.array(labels).astype(np.uint8)
len_data = len(data)
len_labels = len(labels)
assert (len_data == len_labels)
split_len_pos = int(len_data * 0.8)
return (data[:split_len_pos],
labels[:split_len_pos]), (data[split_len_pos:],
labels[split_len_pos:])
def _getPath(self):
"""数据源目录
"""
sources_path = env.get_root_path() + '/cnn_boll/datasources/'
return sources_path
def _get_datas_dir(self):
#cur_dir = os.path.abspath(os.path.dirname(sys.argv[0]) + '/..')
cur_dir = env.get_root_path()
cur_dir += '/datas'
return cur_dir
def _getPath(self):
"""数据源目录
"""
sources_path = 'datasources/'
data_path = env.get_root_path()
return data_path + '/' + sources_path
