python pandas DataFrame: Shuffle 打乱顺序/洗牌

转载：pandas: Shuffle rows/elements of DataFrame/Series

还有一篇更多介绍的：How to Shuffle Pandas Dataframe Rows in Python

You can randomly shuffle rows of pandas.DataFrame and elements of pandas.Series with the sample() method. There are other ways to shuffle, but using the sample() method is convenient because it does not require importing other modules.

• pandas.DataFrame.sample — pandas 1.4.2 documentation

This article describes the following contents.

• Specify frac=1 for sample() to shuffle
• Reset index: ignore_index, reset_index()
• Update original object

In the sample code, the following CSV file is used.

• sample_pandas_normal.csv

• source: pandas_random_sort_shuffle.py

import pandas as pd

df = pd.read_csv('data/src/sample_pandas_normal.csv')
print(df)
#       name  age state  point
# 0    Alice   24    NY     64
# 1      Bob   42    CA     92
# 2  Charlie   18    CA     70
# 3     Dave   68    TX     70
# 4    Ellen   24    CA     88
# 5    Frank   30    NY     57

The example uses pandas.DataFrame, but you can shuffle pandas.Series in the same way.

Note that you can use sort_values() and sort_index() to sort rows according to index or column values. See the following article.

• pandas: Sort DataFrame, Series with sort_values(), sort_index()

Specify frac=1 for sample() to shuffle

See the following article for details of the sample() method.

• pandas: Random sampling from DataFrame with sample()

If the frac parameter is set to 1, all the rows are randomly sampled, equivalent to shuffling the entire row.

print(df.sample(frac=1))
#       name  age state  point
# 2  Charlie   18    CA     70
# 1      Bob   42    CA     92
# 3     Dave   68    TX     70
# 0    Alice   24    NY     64
# 5    Frank   30    NY     57
# 4    Ellen   24    CA     88

You can initialize the random number generator with a fixed seed with the random_state parameter. After initialization with the same seed, they are always shuffled in the same way.

print(df.sample(frac=1, random_state=0))
#       name  age state  point
# 5    Frank   30    NY     57
# 2  Charlie   18    CA     70
# 1      Bob   42    CA     92
# 3     Dave   68    TX     70
# 0    Alice   24    NY     64
# 4    Ellen   24    CA     88

print(df.sample(frac=1, random_state=0))
#       name  age state  point
# 5    Frank   30    NY     57
# 2  Charlie   18    CA     70
# 1      Bob   42    CA     92
# 3     Dave   68    TX     70
# 0    Alice   24    NY     64
# 4    Ellen   24    CA     88

Reset index: ignore_index, reset_index()

If you want to reindex the result (0, 1, ... , n-1), set the ignore_index parameter to True.

print(df.sample(frac=1, ignore_index=True))
#       name  age state  point
# 0    Ellen   24    CA     88
# 1    Frank   30    NY     57
# 2      Bob   42    CA     92
# 3     Dave   68    TX     70
# 4    Alice   24    NY     64
# 5  Charlie   18    CA     70

The ignore_index was added in pandas 1.3.0. For earlier versions, you can use the reset_index() method. Set the drop parameter to True to delete the original index.

• pandas: Reset index of DataFrame, Series with reset_index()

print(df.sample(frac=1).reset_index(drop=True))
#       name  age state  point
# 0      Bob   42    CA     92
# 1     Dave   68    TX     70
# 2    Alice   24    NY     64
# 3  Charlie   18    CA     70
# 4    Frank   30    NY     57
# 5    Ellen   24    CA     88

Update original object

If you want to update the original object, assign the shuffled result to the original object and overwrite it.

df = df.sample(frac=1)
print(df)
#       name  age state  point
# 0    Alice   24    NY     64
# 5    Frank   30    NY     57
# 1      Bob   42    CA     92
# 4    Ellen   24    CA     88
# 3     Dave   68    TX     70
# 2  Charlie   18    CA     70