如何遍历熊猫数据框的列并相应地删除行？ （Jupyter 笔记本）

Question

这是我现在正在使用的数据框：

	Season	Team	W	L	W/L%	Coaches
1	2020-21	Atlanta Hawks*	41	31	0.569	L. Pierce (14-20) N. McMillan (27-11)
2	2019-20	Atlanta Hawks	20	47	0.299	L. Pierce (20-47)
3	2018-19	Atlanta Hawks	29	53	0.354	L. Pierce (29-53)
4	2017-18	Atlanta Hawks	24	58	0.293	M. Budenholzer (24-58)
5	2016-17	Atlanta Hawks*	43	39	0.524	M. Budenholzer (43-39)
6	2015-16	Atlanta Hawks*	48	34	0.585	M. Budenholzer (48-34)
7	2014-15	Atlanta Hawks*	60	22	0.732	M. Budenholzer (60-22)
8	2013-14	Atlanta Hawks*	38	44	0.463	M. Budenholzer (38-44)
9	2012-13	Atlanta Hawks*	44	38	0.537	L. Drew (44-38)
10	2011-12	Atlanta Hawks*	40	26	0.606	L. Drew (40-26)
11	2010-11	Atlanta Hawks*	44	38	0.537	L. Drew (44-38)
12	2009-10	Atlanta Hawks*	53	29	0.646	M. Woodson (53-29)

我基本上只想有教练的名字从一年到下一年不同的行。例如，我会保留第 4 行，因为它紧邻的第 3 行在“教练”列下具有不同的名称，但我会删除第 5、6 和 7 行，因为“教练”列中的名称与所有三行。但是，我想保留第 8 行，因为第 9 行（相邻行）在“教练”中有不同的名称。

我通过读取 csv 文件得到了这个数据框 df = pd.read_csv("hawks.csv") 我想我应该打电话给df.iloc，但我不知道如何遍历每一行并比较列中的值。到目前为止，我只设法在“教练”列中打印字符串值，如下所示：

coaches = df.iloc[0:, 7]
for name in coaches:
    print(name)

但我想知道当我遍历每一行时如何获取存储在“Coachs”列中的值（然后删除不符合我正在寻找的标准的行）。非常感谢！

Answer 1

您可以通过将连续行与shift 进行比较来过滤列：

import pandas as pd

data = [ { "idx": 1, "Season": "2020-21", "Team": "Atlanta Hawks*", "W": 41, "L": 31, "Coaches": "L. Pierce (14-20) N. McMillan (27-11)" }, { "idx": 2, "Season": "2019-20", "Team": "Atlanta Hawks", "W": 20, "L": 47, "Coaches": "L. Pierce (20-47)" }, { "idx": 3, "Season": "2018-19", "Team": "Atlanta Hawks", "W": 29, "L": 53, "Coaches": "L. Pierce (29-53)" }, { "idx": 4, "Season": "2017-18", "Team": "Atlanta Hawks", "W": 24, "L": 58, "Coaches": "M. Budenholzer (24-58)" }, { "idx": 5, "Season": "2016-17", "Team": "Atlanta Hawks*", "W": 43, "L": 39, "Coaches": "M. Budenholzer (43-39)" }, { "idx": 6, "Season": "2015-16", "Team": "Atlanta Hawks*", "W": 48, "L": 34, "Coaches": "M. Budenholzer (48-34)" }, { "idx": 7, "Season": "2014-15", "Team": "Atlanta Hawks*", "W": 60, "L": 22, "Coaches": "M. Budenholzer (60-22)" }, { "idx": 8, "Season": "2013-14", "Team": "Atlanta Hawks*", "W": 38, "L": 44, "Coaches": "M. Budenholzer (38-44)" }, { "idx": 9, "Season": "2012-13", "Team": "Atlanta Hawks*", "W": 44, "L": 38, "Coaches": "L. Drew (44-38)" }, { "idx": 10, "Season": "2011-12", "Team": "Atlanta Hawks*", "W": 40, "L": 26, "Coaches": "L. Drew (40-26)" }, { "idx": 11, "Season": "2010-11", "Team": "Atlanta Hawks*", "W": 44, "L": 38, "Coaches": "L. Drew (44-38)" }, { "idx": 12, "Season": "2009-10", "Team": "Atlanta Hawks*", "W": 53, "L": 29, "Coaches": "M. Woodson (53-29)" } ]
df = pd.DataFrame(data)
df[(df['Coaches'].str.split('(').str[0] != df['Coaches'].shift(1).str.split('(').str[0]) | (df['Coaches'].str.split('(').str[0] != df['Coaches'].shift(-1).str.split('(').str[0])]

输出：

	idx	Season	Team	W	L	Coaches
0	1	2020-21	Atlanta Hawks*	41	31	L. Pierce (14-20) N. McMillan (27-11)
2	3	2018-19	Atlanta Hawks	29	53	L. Pierce (29-53)
3	4	2017-18	Atlanta Hawks	24	58	M. Budenholzer (24-58)
7	8	2013-14	Atlanta Hawks*	38	44	M. Budenholzer (38-44)
8	9	2012-13	Atlanta Hawks*	44	38	L. Drew (44-38)
10	11	2010-11	Atlanta Hawks*	44	38	L. Drew (44-38)
11	12	2009-10	Atlanta Hawks*	53	29	M. Woodson (53-29)

我不确定您是否想将 Pierce/McMillan 视为第一次提到 Pierce，我的回答是这样假设的。如果您想将此计为单独的命中，只需将split('(') 替换为rsplit('(', 1)。

Answer 2

遍历数据框：

it = df.iterrows()  #get iterator
for index, row in it:

    # to delete row
    if cond:
        df.drop([row], inplace=True)

# reset indices if needed
df.reset_index(drop=True, inplace=True)

Answer 3

你可以使用drop_duplicates方法。

comp = re.compile(r'[A-Z].\s\w+')
df['Coaches_Names'] = df['Coaches'].apply(lambda x: ' - '.join(comp.findall(x)))  # New column which includes only coache's names 
df['Team'] = df['Team'].apply(lambda x: ''.join(re.sub('\*', '', x)))   # Dropping '*' characters from name of teams
    
df.drop_duplicates(subset=['Team', 'Coaches_Names'], inplace=True)

print(df)
     Season           Team   W   L   W/L%                                Coaches            Coaches_Names
0   2020-21  Atlanta Hawks  41  31  0.569  L. Pierce (14-20) N. McMillan (27-11)  L. Pierce - N. McMillan
1   2019-20  Atlanta Hawks  20  47  0.299                      L. Pierce (20-47)                L. Pierce
3   2017-18  Atlanta Hawks  24  58  0.293                 M. Budenholzer (24-58)           M. Budenholzer
8   2012-13  Atlanta Hawks  44  38  0.537                        L. Drew (44-38)                  L. Drew
11  2009-10  Atlanta Hawks  53  29  0.646                     M. Woodson (53-29)               M. Woodson

是否删除Coaches_Names 列由您决定。

此外，您可以更喜欢使用keep 参数。它允许您在第一年或去年进行选择。有关更多信息，您可以查看文档。 https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.drop_duplicates.html

编辑

代码被编辑以格式化列，@RJ，Adriaansen