pandas 中的数据修改

Question

我有一个 CSV 文件，其中的行如下所示：

ID,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#,

我可以读入

#!/usr/bin/env python

import pandas as pd
import sys

filename = sys.argv[1]
df = pd.read_csv(filename)

给定一个特定的列，我想按 ID 拆分行，然后输出每个 ID 的均值和标准差。

我的第一个问题是，如何从应该分别为 100 和 0 的“100M”和“0N#”等数字中删除所有 non-numeric 部分。

我还尝试循环遍历相关的 headers 并使用

df[header].replace(regex=True,inplace=True,to_replace=r'\D',value=r'')

如 Pandas DataFrame: remove unwanted parts from strings in a column 中所建议。

但是这会将 98.4 更改为 984。

Answer 1

使用str.extract:

In [356]:
import io
import pandas as pd
t="""ID,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#"""
df = pd.read_csv(io.StringIO(t), header=None)
df

Out[356]:
   0     1     2    3    4    5     6    7    8     9    10    11    12   13  \
0  ID  98.4  100M  55M  65M  75M  100M  75M  65M  100M  98M  100M  100M  92M   

   14   15  
0  0#  0N#  

In [357]:
for col in df.columns[2:]:
    df[col] = df[col].str.extract(r'(\d+)').astype(int)
df

Out[357]:
   0     1    2   3   4   5    6   7   8    9   10   11   12  13  14  15
0  ID  98.4  100  55  65  75  100  75  65  100  98  100  100  92   0   0

如果您有浮点数，则可以使用以下正则表达式：

In [379]:
t="""ID,98.4,100.50M,55.234M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#"""
df = pd.read_csv(io.StringIO(t), header=None)
df

Out[379]:
   0     1        2        3    4    5     6    7    8     9    10    11  \
0  ID  98.4  100.50M  55.234M  65M  75M  100M  75M  65M  100M  98M  100M   

     12   13  14   15  
0  100M  92M  0#  0N#  

In [380]:    
for col in df.columns[2:]:
    df[col] = df[col].str.extract(r'(\d+\.?\d+)').astype(np.float)
df

Out[380]:
   0     1      2       3   4   5    6   7   8    9   10   11   12  13  14  15
0  ID  98.4  100.5  55.234  65  75  100  75  65  100  98  100  100  92 NaN NaN

so (\d+\.?\d+) 查找包含 \d+ 1 位或更多位带 \.? 可选小数点和 \d+ 1 位或更多位小数点后的组

编辑

OK 编辑了我的正则表达式模式：

In [408]:
t="""Name,97.7,0A,0A,65M,0A,100M,5M,75M,100M,90M,90M,99M,90M,0#,0N#"""
df = pd.read_csv(io.StringIO(t), header=None)
df

Out[408]:
     0     1   2   3    4   5     6   7    8     9    10   11   12   13  14  \
0  Name  97.7  0A  0A  65M  0A  100M  5M  75M  100M  90M  90M  99M  90M  0#   

    15  
0  0N#  

In [409]:    
for col in df.columns[2:]:
    df[col] = df[col].str.extract(r'(\d+\.*\d*)').astype(np.float)
df

Out[409]:
     0     1   2   3   4   5    6   7   8    9   10  11  12  13  14  15
0  Name  97.7   0   0  65   0  100   5  75  100  90  90  99  90   0   0

Answer 2

My first problem is, how can I remove all the non-numeric parts from the numbers such as "100M" and "0N#" which should be 100 and 0 respectively.

import re
df = pd.read_csv(yourfile, header=None)
df.columns = ['ID'] + list(df.columns)[1:]
df = df.stack().apply(lambda v: re.sub('[^0-9]','', v) 
                 if isinstance(v, str) else v).astype(float).unstack()
df.groupby('ID').agg(['std', 'mean'])

此处 .stack() 将数据帧转换为系列，.apply() 为每个值调用 lambda，re.sub() 删除任何非数字字符，.astype() 转换为数字unstack() 将系列转换回数据框。这对整数和浮点数同样有效。

Given a particular column, I would like to split the rows by ID and then output the mean and standard deviation for each ID.

 # for all columns
 df.groupby('ID').agg(['std', 'mean'])
 # for specific column
 df.groupby('ID')['<colname>'].agg(['std', 'mean'])

示例中使用的数据如下：

from StringIO import StringIO
s="""
1,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#,
1,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#,
2,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#,
2,98.4,100M,55M,65M,75M,100M,75M,65M,100M,98M,100M,100M,92M,0#,0N#,
"""
yourfile = StringIO(s)

pandas 中的数据修改

Data munging in pandas

python

pandas

dataframe

data-manipulation

data-munging