迭代数据框并根据一列的值在具有前一行值的新列中执行操作

Posted 2023-02-18

【中文标题】迭代数据框并根据一列的值在具有前一行值的新列中执行操作

【英文标题】：iterrate over dataframe and based on the value of one column do operations in a new column with previous row's value

【发布时间】：2021-11-23 18:01:10

【问题描述】：

我对他们的行为有少量的股票价格。我想计算拆分后股票的调整所有权数量（即，如果您拥有 1000 股并且股票有 2-1 拆分，那么您的所有权变为 2000 股）。我想遍历“Stock Splits”列，如果值！= 0，则将“所有权”与“Stock Splits”相乘，否则保持拆分前的最后一个数量。我尝试了很多方法，但我不确定我哪里出错了 - 我确实认为逻辑是错误的，但不知道如何解决它。

import yfinance as yf
aapl = yf.Ticker("AAPL")
hist = aapl.history(start="2014-06-01")
hist["ownership"] = 1000


    Open    High    Low Close   Volume  Dividends   Stock Splits    ownership
Date                                
2014-06-02  20.338966   20.366877   19.971301   20.168608   369350800   0.0 0.0 1000
2014-06-03  20.162511   20.492319   20.155774   20.453819   292709200   0.0 0.0 1000
2014-06-04  20.450610   20.785872   20.407940   20.687378   335482000   0.0 0.0 1000
2014-06-05  20.731655   20.833356   20.616479   20.768549   303805600   0.0 0.0 1000
2014-06-06  20.850357   20.893990   20.676150   20.711439   349938400   0.0 0.0 1000 

我的代码如下：

 hist.loc[hist['Stock Splits']==0,'ownerAdj'] = hist['ownership'].shift(1)
hist.loc[hist['Stock Splits']!=0,'ownerAdj'] = hist['ownership'].shift(1) * hist['Stock Splits']

但是我并不总是得到正确的数字，如下例所示，在 2014 年 6 月 9 日 aapl 已拆分（7 比 1），因此从 2014 年 6 月 9 日到下一个日期的结果应该是 7000另一个拆分是 2020-08-31，但拆分后我要回 1000

Date    Open    High    Low Close   Volume  Dividends   Stock Splits    ownership   ownerAdj
0   2014-06-02  20.338964   20.366875   19.971299   20.168606   369350800   0.0 0.0 1000    NaN
1   2014-06-03  20.162515   20.492323   20.155778   20.453823   292709200   0.0 0.0 1000    1000.0
2   2014-06-04  20.450608   20.785870   20.407938   20.687376   335482000   0.0 0.0 1000    1000.0
3   2014-06-05  20.731645   20.833346   20.616470   20.768539   303805600   0.0 0.0 1000    1000.0
4   2014-06-06  20.850359   20.893992   20.676152   20.711441   349938400   0.0 0.0 1000    1000.0
5   2014-06-09  20.818268   21.083269   20.604921   21.042845   301660000   0.0 7.0 1000    7000.0
6   2014-06-10  21.274162   21.346027   21.013652   21.166365   251108000   0.0 0.0 1000    1000.0
7   2014-06-11  21.139424   21.280908   20.991204   21.078789   182724000   0.0 0.0 1000    1000.0

我尝试运行循环，但出现错误：

for i, row in hist.iterrows():
    if row["Stock Splits"] == 0:
        row["ownerAdj"] = row["ownership"].shift(1)
    elif row["Stock Splits"] != 0:
        row["ownerAdj"] = row["ownership"].shift(1) * row["Stock Splits"]

 ---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-51-2d94c5e86953> in <module>
      1 for i, row in hist.iterrows():
      2     if row["Stock Splits"] == 0:
----> 3         row["adjust2"] = row["ownership"].shift(1)
      4     elif row["Stock Splits"] != 0:
      5         row["adjust2"] = row["ownership"].shift(1) * row["Stock Splits"]

AttributeError: 'numpy.float64' object has no attribute 'shift'

【参考方案1】：

你可以做这个矢量化

hist['ownership'] = 1000 * np.cumprod(np.maximum(hist["Stock Splits"], 1))

部分：

# No split can be expressed as a 1.0 split (You get 1 for every 1).
# Assumes you don't have negative splits.
adj_split = np.maximum(hist["Stock Splits"], 1)  

# The multiple of the initial ownership at each day compared to the first.
cumsplit = np.cumprod(adj_split)

initial_ownership = 1000
hist["ownership"] = cumsplit * initial_ownership

【讨论】：

感谢您的回复。运行 hist['ownership2'] = 1000 * np.cumprod(np.maximum(hist["Stock Splits"],1)

时出现 EOF 错误

对不起，忘记了结束括号:)