Pandas遍历Series遇KeyError：如何检查NumPy浮点值是否大于0？

我有一个名为df1的DataFrame，结构如下：

Open    Close         DiffMa    %Percentage  Open   Close   DiffMa2  %Percentage2
2022-11-04 13:30:00-04:00   42.099998   42.224998   -0.135001   0.296912    13.9586 14.0150 -0.06584    0.404054
2022-11-04 14:30:00-04:00   42.220001   42.330002   0.028002    0.260541    14.0150 14.0550  0.01816        0.285408
2022-11-04 15:30:00-04:00   42.320000   42.630001   0.311001    0.732517    14.0600 14.1100  0.07716        0.355613
2022-11-07 09:30:00-05:00   43.049999   42.294998   -0.019002   -1.753777   14.3200 14.0299 -0.00308    -2.025839
2022-11-07 10:30:00-05:00   42.299999   42.195000   -0.140000   -0.248226   14.0300 13.9865 -0.05278    -0.310050

df1['%Percentage'][0]的输出为：0.2969121247755807

我需要统计%Percentage和%Percentage2两列同时为正或同时为负的次数，但编写循环时触发KeyError，简化后的循环代码：

countera = 0
counterb = 0

for i in df1['%Percentage']:
    if df1['%Percentage'][i] > 0:
       countera = 0

报错信息：

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-10-cc9f6f8fa7d5> in <module>
      4 
      5 for i in df1['%Percentage']:
----> 6     if df1['%Percentage'][i] > 0:
      7         countera += 1
      8 

~/opt/anaconda3/envs/StockPredictionGameification/lib/python3.6/site-packages/pandas/core/series.py in __getitem__(self, key)
    880 
    881         elif key_is_scalar:
--> 882             return self._get_value(key)
    883 
    884         if is_hashable(key):

~/opt/anaconda3/envs/StockPredictionGameification/lib/python3.6/site-packages/pandas/core/series.py in _get_value(self, label, takeable)
    988 
    989         # Similar to Index.get_value, but we do not fall back to positional
--> 990         loc = self.index.get_loc(label)
    991         return self.index._get_values_for_loc(self, loc, label)
    992 

~/opt/anaconda3/envs/StockPredictionGameification/lib/python3.6/site-packages/pandas/core/indexes/datetimes.py in get_loc(self, key, method, tolerance)
    620         else:
    621             # unrecognized type
--> 622             raise KeyError(key)
    623 
    624         try:

KeyError: 0.2969121247755807

我猜测是浮点数精度问题，但不确定具体原因，需要问题解析和解决办法。

你遇到的KeyError和浮点数精度无关，核心问题是循环变量的理解错误：

• for i in df1['%Percentage']遍历的是该列的数值（比如第一个i是0.2969121247755807）
• 而df1['%Percentage'][i]是试图用这个数值作为索引去查找数据，但你的DataFrame索引是时间戳（比如2022-11-04 13:30:00-04:00），时间戳索引里不存在0.2969121247755807这个值，因此触发KeyError。

方法1：修复循环逻辑 #

如果一定要用循环，可以遍历索引或同时遍历两列的数值对：

方式A：遍历索引 #

counter_both_pos = 0
counter_both_neg = 0

for idx in df1.index:
    p1 = df1['%Percentage'][idx]
    p2 = df1['%Percentage2'][idx]
    if p1 > 0 and p2 > 0:
        counter_both_pos += 1
    elif p1 < 0 and p2 < 0:
        counter_both_neg += 1

print(f"同时为正次数：{counter_both_pos}")
print(f"同时为负次数：{counter_both_neg}")

方式B：直接遍历两列的数值对 #

counter_both_pos = 0
counter_both_neg = 0

for p1, p2 in zip(df1['%Percentage'], df1['%Percentage2']):
    if p1 > 0 and p2 > 0:
        counter_both_pos += 1
    elif p1 < 0 and p2 < 0:
        counter_both_neg += 1

print(f"同时为正次数：{counter_both_pos}")
print(f"同时为负次数：{counter_both_neg}")

方法2：Pandas向量化操作（推荐） #

Pandas的核心优势是向量化运算，比循环高效得多，无需遍历每一行：

# 计算同时为正的布尔序列，sum()统计True的数量
both_pos = ((df1['%Percentage'] > 0) & (df1['%Percentage2'] > 0)).sum()
# 计算同时为负的布尔序列
both_neg = ((df1['%Percentage'] < 0) & (df1['%Percentage2'] < 0)).sum()

print(f"同时为正次数：{both_pos}")
print(f"同时为负次数：{both_neg}")

这种方式代码更简洁，执行效率远高于循环，尤其当DataFrame数据量较大时差距明显。

内容的提问来源于stack exchange，提问作者Niko