拆分熊猫日期时间索引以创建分类变量

Question

我必须从 pandas 日期时间索引中创建一个分类变量，并为它寻找一种 Python 方式。

到目前为止，我只是遍历了所有索引并做了一堆 if-else。我尝试使用 (Adding a new pandas column with mapped value from a dictionary) 的灵感，使用 lambda if else 函数的字典，并使用 map 来创建分类函数，但它不起作用

date_series = pd.date_range(start = '2010-12-31', end = '2018-12-31', freq = 'M')

regime_splitter = {lambda x : x < '2012' : 'before 2012' , lambda x : x>= '2012' and x < '2014': '2012 - 2014', lambda x : x>= '2014' : 'after 2014'}

date_series.map(regime_splitter)

预期结果

         date              regime
0  2010-12-31         before 2012
1  2013-05-31  between 2012, 2014
2  2018-12-31          after 2014

Answer 1

import pandas as pd
data_series = pd.date_range(start='2010-12-31', end='2018-12-31', freq='M')
df = pd.DataFrame(data_series, columns=['Dates'])
 
def regime_splitter(value):
    if value < pd.to_datetime('2012-01-01'):
        return 'before 2012'
    elif value > pd.to_datetime('2014-12-31'):
        return'After 2014'
    else:
        return 'Between 2012, 2014'
 
df['regime_splitter'] = df['Dates'].apply(regime_splitter)
 
df.head(15)
 
Dates     regime_splitter
0              2010-12-31           before 2012
1              2011-01-31           before 2012
2              2011-02-28           before 2012
3              2011-03-31           before 2012
4              2011-04-30           before 2012
5              2011-05-31           before 2012
6              2011-06-30           before 2012
7              2011-07-31           before 2012
8              2011-08-31           before 2012
9              2011-09-30           before 2012
10           2011-10-31           before 2012
11           2011-11-30           before 2012
12           2011-12-31           before 2012
13           2012-01-31           Between 2012, 2014
14           2012-02-29           Between 2012, 2014

Answer 2

如果需要添加/删除更多组，请使用 cut 和 DatetimeIndex.year 解决方案：

a = pd.cut(date_series.year, 
       bins=[-np.inf, 2012, 2014, np.inf], 
       labels=['before 2012','2012 - 2014','after 2014'])
print (a.value_counts())
before 2012    25
2012 - 2014    24
after 2014     48
dtype: int64

numpy.select 的另一个解决方案：

x = date_series.year
a = np.select([x <= 2012, x>= 2014], ['before 2012','after 2014'], '2012 - 2014')

print (pd.Series(a).value_counts())
after 2014     60
before 2012    25
2012 - 2014    12
dtype: int64

你的解决方案应该改成嵌套if-else，但是如果数据量大的话应该会很慢：

regime_splitter = (lambda x: 'before 2012' if x <= 2012 else 
                             ('2012 - 2014' if x>= 2012 and x <= 2014 else 'after 2014'))

a = date_series.year.map(regime_splitter)
print (a.value_counts())
after 2014     48
before 2012    25
2012 - 2014    24
dtype: int64