我不熟悉使用熊猫，所以请指出我在代码中可能遗漏的所有最佳做法。我编写了以下python代码，下面的代码从Yahoo Finance下载股票行情收录器数据，仅保留AAPL和MSFT的每日调整后收盘价。然后，我在数据框中创建了一个新列，其中包含AAPL的每日调整后收盘价除以MSFT的每日调整后收盘价（A），以及另一列，即同一计算的100天移动平均线（B）。后第二列是（A）/（B），最后一列是倒数第二天的每日变化百分比。
import pandas as pd
pd.core.common.is_list_like = pd.api.types.is_list_like
from pandas_datareader import data as pdr
import fix_yahoo_finance as yf
yf.pdr_override()
import matplotlib.pyplot as plt
import pandas as pd
# Define the instruments to download. We would like to see Apple, and Microsoft
tickers = ['AAPL', 'MSFT']
# We would like all available data from 01/01/2000 until 12/31/2016.
start_date = '2010-01-01'
end_date = '2016-12-31'
# User pandas_reader.data.DataReader to load the desired data. As simple as that.
df = pdr.get_data_yahoo(tickers, start_date, end_date)['Adj Close']
df["AAPL/MSFT"] = df['AAPL']/df['MSFT']
df["100Day-MA APPL/MSFT"] = (df['AAPL']/df['MSFT']).rolling(window=100).mean()
df["AAPL/MSFT / MA"] = df["AAPL/MSFT"]/df["100Day-MA APPL/MSFT"]
df["% change AAPL/MSFT / MA"] = df["AAPL/MSFT / MA"].pct_change()
print(df.tail(9))
输出如下。我的问题是我该如何设置条件声明，以查看每日价格趋势，并根据这些信号发出买/卖信号。例如，df的最后一栏表示％的变化已连续5天增加。或者，倒数第二列已连续增加5天，但仍低于1。任何见解将不胜感激：）
AAPL MSFT ... AAPL/MSFT / MA % change AAPL/MSFT / MA
Date ...
2016-12-19 113.490685 61.435783 ... 0.965480 -0.014949
2016-12-20 113.792313 61.358532 ... 0.969278 0.003933
2016-12-21 113.899338 61.358532 ... 0.970376 0.001133
2016-12-22 113.150139 61.368191 ... 0.963949 -0.006623
2016-12-23 113.373917 61.068832 ... 0.970687 0.006990
2016-12-27 114.093948 61.107452 ... 0.976266 0.005748
2016-12-28 113.607445 60.827412 ... 0.976658 0.000401
2016-12-29 113.578247 60.740505 ... 0.977934 0.001307
2016-12-30 112.692825 60.006588 ... 0.982284 0.004448
答案 0 :(得分：1)
您的问题要问很多。我将向您展示如何找到连续的增长。
从数据框的简化版本开始：
Date AAPL MSFT change MA
0 2016-12-19 113.490685 61.435783 0.965480 -0.014949
1 2016-12-20 113.792313 61.358532 0.969278 0.003933
2 2016-12-21 113.899338 61.358532 0.970376 0.001133
3 2016-12-22 113.150139 61.368191 0.963949 -0.006623
4 2016-12-23 113.373917 61.068832 0.970687 0.006990
5 2016-12-27 114.093948 61.107452 0.976266 0.005748
6 2016-12-28 113.607445 60.827412 0.976658 0.000401
7 2016-12-29 113.578247 60.740505 0.977934 0.001307
8 2016-12-30 112.692825 60.006588 0.982284 0.004448
我们将询问change列是否比前一天有所增加：
df['inc'] = df['change'] > df['change'].shift()
这给我们留下了
Date AAPL MSFT change MA inc
0 2016-12-19 113.490685 61.435783 0.965480 -0.014949 False
1 2016-12-20 113.792313 61.358532 0.969278 0.003933 True
2 2016-12-21 113.899338 61.358532 0.970376 0.001133 True
3 2016-12-22 113.150139 61.368191 0.963949 -0.006623 False
4 2016-12-23 113.373917 61.068832 0.970687 0.006990 True
5 2016-12-27 114.093948 61.107452 0.976266 0.005748 True
6 2016-12-28 113.607445 60.827412 0.976658 0.000401 True
7 2016-12-29 113.578247 60.740505 0.977934 0.001307 True
8 2016-12-30 112.692825 60.006588 0.982284 0.004448 True
然后，我们要求在为期5天的跟踪窗口中选择最小值；仅当所有5天均为True时，该值将为True：
df['inc5'] = df['inc'].rolling(5).min().fillna(0).astype('bool')
因此我们有：
Date AAPL MSFT change MA inc inc5
0 2016-12-19 113.490685 61.435783 0.965480 -0.014949 False False
1 2016-12-20 113.792313 61.358532 0.969278 0.003933 True False
2 2016-12-21 113.899338 61.358532 0.970376 0.001133 True False
3 2016-12-22 113.150139 61.368191 0.963949 -0.006623 False False
4 2016-12-23 113.373917 61.068832 0.970687 0.006990 True False
5 2016-12-27 114.093948 61.107452 0.976266 0.005748 True False
6 2016-12-28 113.607445 60.827412 0.976658 0.000401 True False
7 2016-12-29 113.578247 60.740505 0.977934 0.001307 True False
8 2016-12-30 112.692825 60.006588 0.982284 0.004448 True True
答案 1 :(得分：0)
一种方法是计算差异，如果差异大于0，则创建一个列，否则为0。然后对最后一列进行滚动总和，如果得到5，则执行某项操作。 / p>
df['diff'] = df["AAPL/MSFT / MA"].diff()
df.loc[df['diff'] > 0, 'sign'] = 1
df.loc[df['diff'] <= 0, 'sign'] = 0
df['sum'] = df['sign'].rolling(5).sum()
if len(df.loc[df['sum'] == 5, 'sum']) > 0:
all_your_stocks_are_belong_to_us()
这假定所有连续的天都存在并且时间序列中没有空洞。如果不是这种情况，则需要使用更多的技巧。