def get_sd(id, label_start_date, fea_delta = 240):
# delta 指的是 确定label；的时候 是一个星期的涨幅还是什么
# fea delta 指的是， 我要准备多久的数据
## 获取分笔数据
# get all time ticks
# 创建一个日期列表
ticks = None
date_list = dt_tool.dt_range(label_start_date, -fea_delta)
for date in date_list:
try:
tick = ts.get_tick_data(id, date)
except Exception, e:
print e
continue
if tick is None:
continue
tick.type = tick.type.apply(lambda x : type2id(x))
ft = tick.sort('amount', ascending=False).head(10).reset_index().drop(['index', 'time', 'change'], 1).stack().reset_index()
ft.index = ft.level_0.map(str) + '_' + ft.level_1
fT = ft.drop(['level_0', 'level_1'], 1).T
fT['date'] = dt_tool.format(date)
if ticks is None:
ticks = fT
else:
ticks = ticks.append(fT)
ticks.to_csv('data/ticks.csv', index=None)
def get_tick_feature(id, end_str, delta):
if isinstance(end_str, str):
end = dt.datetime.strptime(end_str, '%Y-%m-%d')
date_list = [end - dt.timedelta(days=x) for x in range(0, 90)]
ticks = None
for date in date_list:
date_str = dt.datetime.strftime(date, '%Y-%m-%d')
try:
tick = ts.get_tick_data(id, date_str)
except:
continue
if tick is None:
continue
ft = tick.sort('amount', ascending=False).head(10).reset_index().drop(['index', 'time'], 1).stack().reset_index()
ft['fea'] = ft.level_0.map(str) + '_' + ft.level_1
fea_pd = ft.drop(['level_0', 'level_1'], 1)
fea_pd.index = fea_pd.fea
fT = fea_pd.T
fT['date'] = dt_tool.format(date_str)
if ticks is None:
ticks = fT
else:
ticks = ticks.append(fT)
return ticks
def get_sd(id):
try:
hist = ts.get_h_data(id, autype='qfq', start='2005-06-10')
except:
return None
ls = hist.index.format()
hist.index = [ dt_tool.format(x) for x in ls]
fea_tick = get_tick_feature(id, '2015-06-10', 1500)
print 'fea', fea_tick.columns
print 'hist', hist.columns
hist.to_csv('hist.csv')
fea_tick.to_csv('tick.csv')
h = hist.merge(fea_tick, how='left', left_index=True, right_on='date')
h.to_csv('merged.csv')
s = h.stack()
spd = pd.DataFrame(s)
sd = spd.reset_index()
sd.columns = ('date', 'type', 'value')
return sd
fT['date'] = dt_tool.format(date)
if ticks is None:
ticks = fT
else:
ticks = ticks.append(fT)
ticks.to_csv('data/ticks.csv', index=None)
# 获取历史数据矩阵
# 将数据stack为series, 方便处理
try:
#hist = ts.get_h_data(id, autype='qfq', start='2005-06-10')
hist = ts.get_hist_data(id)
except Exception, e:
print e
ls = hist.index.format()
hist['date'] = [dt_tool.format(x) for x in ls]
hist.to_csv('data/hist.csv', index=None)
if hist is None or ticks is None:
return None
# 聚合两份数据
h = hist.merge(ticks, how='left', on='date')
h.to_csv('data/merged.csv', index=None)
h.index = h.date
h = h.drop('date', 1)
s = h.stack()
spd = pd.DataFrame(s)
sd = spd.reset_index()
sd.columns = ('date', 'type', 'value')
sd.to_csv('data/sd.csv', index=None)
