如何在 pandas.apply 函数中索引和使用当前和先前的列值来计算下一列值？

Question

我正在尝试构建一个函数，该函数使用 pandas 数据帧中的收盘价和 ATR 值计算追踪止损值。

供参考的dataframe如下：

            High    Low     Open    Close       ATR 
Date                                
2020-06-01  5.88    5.67    5.73    5.87        0.210000    
2020-06-02  6.00    5.83    5.96    5.90        0.207143    
2020-06-03  6.27    5.92    5.99    6.19        0.218776    
2020-06-04  6.58    6.12    6.20    6.57        0.236006    
2020-06-05  7.50    7.02    7.24    7.34        0.285577    
2020-06-08  7.74    7.37    7.53    7.53        0.293750    
2020-06-09  7.44    7.05    7.22    7.24        0.307053    
2020-06-10  7.34    6.77    7.33    6.81        0.325835    
2020-06-11  6.46    6.04    6.07    6.13        0.357561

我想要的样子：


            High    Low     Open    Close       ATR         ATR_TS
Date                                
2020-06-01  5.88    5.67    5.73    5.87        0.210000    5.135000
2020-06-02  6.00    5.83    5.96    5.90        0.207143    5.175000
2020-06-03  6.27    5.92    5.99    6.19        0.218776    5.424286
2020-06-04  6.58    6.12    6.20    6.57        0.236006    5.743980
2020-06-05  7.50    7.02    7.24    7.34        0.285577    6.340481
2020-06-08  7.74    7.37    7.53    7.53        0.293750    6.501876
2020-06-09  7.44    7.05    7.22    7.24        0.307053    6.501876
2020-06-10  7.34    6.77    7.33    6.81        0.325835    6.501876
2020-06-11  6.46    6.04    6.07    6.13        0.357561    7.381464

我的 pseudo-function/logic 目前是这样的：

def atr_ts(close, atr):
    bigatr = atr*3.5
    buysell = 1 
    stop[i-1] = 0  
    
    if buysell > 0:
        stop = close - bigatr
        stop = max(stop, stop[i-1])
        if close < stop:
            stop = close + bigatr
            buysell = -1

    elif buysell < 0:
        stop = close + bigatr
        stop = min(stop, stop[i-1])
        if close > stop:
            stop = close - bigatr
            buysell = 1
        
            
    return stop

df['ATR_TS'] = df.apply(lambda col: atr_ts(col['Close'], col['ATR']), axis = 1)

所以我的问题是，如何索引这个函数计算的前一个停止值（ATR_TS）来计算下一个停止值，第一个停止值是0？如果有人在不使用 pandas.apply 的情况下看到更好的解决此问题的方法，也请分享。

总体而言，我是编程新手，所以如果不清楚，我深表歉意。

非常感谢。

Answer 1

这不是最终的解决方案，因为我很困惑为什么 ATR_TS = 7.381464 的最后一个值，尽管我知道您是如何计算该值的。我创建了许多列来可视化使用 .shift() 和 .cumsum() 进行某些计算的“pandonic”方式进行 row-wise 比较。请查看列和我的屏幕截图并解释如何获取最后一个值，但在最终解决方案中绝对不需要所有这些列：

df['bigatr'] = (df['ATR'] * 3.5)
df['Stop1a'] = df['Close'] - (df['ATR'] * 3.5)
df['Stop2a'] = df.shift()['Close'] - (df.shift()['ATR'] * 3.5)
df['Stop3a'] = df[['Stop1a','Stop2a']].max(axis=1)
df['Stop1b'] = df['Close'] + (df['ATR'] * 3.5)
df['Stop2b'] = df.shift()['Close'] + (df.shift()['ATR'] * 3.5)
df['Stop3b'] = df[['Stop1b','Stop2b']].min(axis=1)
df['cuma'] = (df['Stop1a'] > df.shift()['Stop1a']).cumsum()
df['cumb'] = (df['Stop1b'] < df.shift()['Stop1b']).cumsum()
df['ATR_TSa'] = df.groupby((df['Stop1a'] > df.shift()['Stop1a']).cumsum())['Stop1a'].transform('first')
df['ATR_TSb'] = df.groupby((df['Stop1b'] < df.shift()['Stop1b']).cumsum())['Stop1b'].transform('first')
df

如您所见，最终的解决方案是最后一行red中圈出的['ATR_TSa']值和blue中圈出的df['ATR_TSb']值。

编辑 #1 - 根据 OP 的评论，解决上述问题的最终逻辑是添加：

 df['ATR_TS'] = np.where((df['Close'] < df['ATR_TSa']), df['ATR_TSb'], df['ATR_TSa'])

现在，在下面，我将提供一个更简洁的解决方案：

df['Stop1a'] = df['Close'] - (df['ATR'] * 3.5)
df['Stop1b'] = df['Close'] + (df['ATR'] * 3.5)
a = df.groupby((df['Stop1a'] > df.shift()['Stop1a']).cumsum())['Stop1a'].transform('first')
b = df.groupby((df['Stop1b'] <= df.shift()['Stop1b']).cumsum())['Stop1b'].transform('first')
df['ATR_TS'] = np.where((df['Close'] < a), b, a)
df = df.drop(['Stop1a','Stop1b'], axis=1)
df


Out[1]: 
         Date  High   Low  Open  Close       ATR    ATR_TS
0  2020-06-01  5.88  5.67  5.73   5.87  0.210000  5.135000
1  2020-06-02  6.00  5.83  5.96   5.90  0.207143  5.175000
2  2020-06-03  6.27  5.92  5.99   6.19  0.218776  5.424284
3  2020-06-04  6.58  6.12  6.20   6.57  0.236006  5.743979
4  2020-06-05  7.50  7.02  7.24   7.34  0.285577  6.340480
5  2020-06-08  7.74  7.37  7.53   7.53  0.293750  6.501875
6  2020-06-09  7.44  7.05  7.22   7.24  0.307053  6.501875
7  2020-06-10  7.34  6.77  7.33   6.81  0.325835  6.501875
8  2020-06-11  6.46  6.04  6.07   6.13  0.357561  7.381463

如何在 pandas.apply 函数中索引和使用当前和先前的列值来计算下一列值？

How do you index and use current and previous column values to calculate the next column value in a pandas.apply function?

python

python-3.x

pandas

quantitative-finance