Python/Pandas - 使用第一个/最后一个函数聚合数据帧而不分组 [英] Python/Pandas - Aggregating dataframe with first/last function without grouping

问题描述

我正在尝试使用 Pandas 聚合整个数据框，而不按任何内容分组.

I am trying to aggregate an entire dataframe using pandas, without grouping by anything.

我确实需要为不同的列使用不同的函数，所以我使用的是字典，但是将first"或last"作为聚合函数传递会引发 ValueError: no results，而其他的，例如 'min'/'max'/'mean' 没问题.

I do need different functions for different columns so I'm using a dictionary, however passing 'first' or 'last' as aggregation functions throws a ValueError: no results, while others such as 'min'/'max'/'mean' give no problem.

这是代码的简化.

df = pd.DataFrame({'Col1':[1,2,3,4], 'Col2':[5,6,7,8], 'Col3':[9,10,11,12]})

func = {col: ['first', 'last'] if col in ['Col1']
             else ['first', 'last', 'mean'] if col in ['Col2']
             else 'mean' for col in df.columns}

result = df.agg(func)

使用

result = df.groupby(lambda _ : True).agg(func)

完成这项工作但速度很慢，我认为是由于 groupby.该数据帧已经是无法进一步分组的更大数据帧的子集.

does the job but is quite slow, I assume due to the groupby. The dataframe is already a subset of a larger dataframe that cannot be further grouped.

我有数百个列，我无法单独聚合它们.

I have hundreds of columns, I cannot aggregate them individually.

是否有另一种方法可以比分组更快/更有效地获取第一行和最后一行以及不同的聚合?

Is there another way to obtain the first and last row, as well as different aggregations, in a faster/more efficient way than grouping?

对于这样的示例数据帧

   Col1  Col2  Col3
0     1     5     9
1     2     6    10
2     3     7    11
3     4     8    12

输出应该是

      Col1       Col2            Col3
     first last first last mean  mean
True     1    4     5    8  6.5  10.5

正如原始 groupby 函数所做的那样，不应删除空值/列.

As the original groupby functions would do, no null values/columns should be removed.

推荐答案

更新:

df = pd.DataFrame({'Col1':[1,2,3,4], 'Col2':[5,6,7,8], 'Col3':[9,10,11,12]})
group_1 = ['Col1']
group_2 = ['col2']
func = {col:[fvalue, lvalue] if col in group_1
         else [fvalue, lvalue, 'mean'] if col in group_2
         else 'mean' for col in df.columns}

df.agg(func).unstack().to_frame().dropna().T

输出:

    Col1        Col2  Col3
  fvalue lvalue mean  mean
0    1.0    4.0  6.5  10.5

让我们看看在不使用 groupby 的情况下使用自定义函数是否会有所帮助:

Let's see if using custom functions without using groupby will help things a little:

def fvalue(x):
    return x.iloc[0]

def lvalue(x):
    return x.iloc[-1]

func = {col:[fvalue, lvalue] if col in group_1
         else [fvalue, lvalue, 'mean'] if col in group_2
         else 'mean' for col in df.columns}

df.agg(func)

这篇关于Python/Pandas - 使用第一个/最后一个函数聚合数据帧而不分组的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！