我正在嘗試計算金融時間序列數據的滾動指標。我想使用循環方法來模擬實時數據的測試。在 itertuples 循環中計算這些滾動指標的最有效方法是什么？示例數據：DateTime Bid 2006-01-03 00:01:07.588 0.85208 2006-01-03 00:01:08.654 0.85213 2006-01-03 00:01:08.859 0.85212 2006-01-03 00:01:11.472 0.85215 2006-01-03 00:01:12.002 0.85218 ... ... 2020-03-15 23:59:57.150 0.85178 2020-03-15 23:59:57.300 0.85179 2020-03-15 23:59:58.233 0.85179 2020-03-15 23:59:58.366 0.85178 2020-03-15 23:59:58.595 0.85179 到目前為止我有代碼。df = pd.read_hdf(r"F:\Market Data\2020.3.15 FXAUDCAD-TICK-NoSession.h5")df = df.set_index(pd.DatetimeIndex(df['DateTime']))df = df.drop(columns=['DateTime'])Rolling_Metric = []for row in df.itertuples(): ?
1 回答
繁花不似錦
TA貢獻1851條經驗 獲得超4個贊
我認為 pandas 不適合解決這個問題。
如果你有像api這樣的數據生成源，這里由生成器模擬
import numpy as np
def gen():
while True:
yield np.random.rand((np.random.randint(1,10)))
輸出不同大小的數據數組
for i in islice(gen(), 4):
print(i)
輸出
[0.1591485]
[0.40462191 0.32921298 0.64704824 0.9433797 0.44754502 0.47600713
0.66130654]
[0.45582976 0.37764161 0.47205139 0.32354448 0.06795233 0.47943393
0.13395702]
[0.0967848]
例如，您可以使用 10 個樣本的窗口來計算滾動測量值
import time
from itertools import islice
data = np.array([])
for new_data in islice(gen(), 5): # get data
for elem in new_data: # iterate through new data
data = np.concatenate((data, [elem])) # add new data row by row
print(data[-10:].mean()) # get mean of last 10 observations
time.sleep(.5)
輸出
0.8251054981003462
0.5154331864262989
0.5677470477572374
0.6084844147856047
0.6532425615231122
0.6663683916931894
0.6768810511903373
0.6098697771903554
0.5976415974047367
0.5442112622703545
0.556721858529291
0.5851107975154073
0.6129548571751687
0.5519507890295304
0.47809901125252807
0.457599927037135
0.47739535574047764
0.5135494376774083
0.5620825459637069
0.5914086396034781
0.5554789093102113
0.6042456773490161
0.5860524867501515
0.6218627945520632
0.6509948271807725
0.6693775700674035
0.6657165569407465
0.6825455302579173
0.609296884720923
0.6708821735456445
添加回答
舉報
0/150
提交
取消