创建数据框并将列设置为日期时间？

Question

我想创建一个 pandas 数据框 df，例如：

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
        "value": [10, 20, 16, 31, 56],
    }
)

我想在创建数据框时将“日期”列指定为 dtype=datetime64[ns]，而不是之后。

所以不是这样的：

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
        "value": [10, 20, 16, 31, 56],
    }
)
df["date"] = pd.to_datetime(df["date"])

但是像这样：

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": pd.Series(
            ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
            dtype=np.datetime64,
        ),
        "value": [10, 20, 16, 31, 56],
    }
)

但这给出了错误：

ValueError: 'datetime64' dtype 没有单位。请改为传入 'datetime64[ns]'。

这样做

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": pd.Series(
            ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
            dtype=np.datetime64[ns],
        ),
        "value": [10, 20, 16, 31, 56],
    }
) 

我收到此错误：

NameError: 名称“ns”未定义

那么如何将特定列设置为“datetime64[ns]”类型？

Answer 1

你可以在dict中使用pd.to_datetime，如下：

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": pd.to_datetime(["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"]),
        "value": [10, 20, 16, 31, 56],
    }
)

date 列采用datetime64[ns] 格式，您可以通过df.info() 看到

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5 entries, 0 to 4
Data columns (total 3 columns):
 #   Column  Non-Null Count  Dtype         
---  ------  --------------  -----         
 0   group   5 non-null      object        
 1   date    5 non-null      datetime64[ns]
 2   value   5 non-null      int64         
dtypes: datetime64[ns](1), int64(1), object(1)
memory usage: 248.0+ bytes

Answer 2

使用dtype='datetime64[ns]':

df=pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": pd.Series(
            ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
            dtype='datetime64[ns]',
        ),
        "value": [10, 20, 16, 31, 56],
    }
)

或

使用astype():

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
        "value": [10, 20, 16, 31, 56],
    }
)
df['date']=df['date'].astype("Datetime64")

或

通过pd.Timestamp():

df = pd.DataFrame(
    {
        "group": ["A", "A", "A", "A", "A"],
        "date": list(map(pd.Timestamp,["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"])),
        "value": [10, 20, 16, 31, 56],
    }
)

Answer 3

我不知道为什么在创建数据框后转换日期会出现问题？您可以将 df 创建与 astype 链接起来：

df = pd.DataFrame(
    {"group": ["A", "A", "A", "A", "A"],
     "date": ["2020-01-02", "2020-01-13", "2020-02-01", "2020-02-23", "2020-03-05"],
     "value": [10, 20, 16, 31, 56],
    }).astype({'date': 'datetime64'})

>>> df.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5 entries, 0 to 4
Data columns (total 3 columns):
 #   Column  Non-Null Count  Dtype
---  ------  --------------  -----
 0   group   5 non-null      object
 1   date    5 non-null      datetime64[ns]
 2   value   5 non-null      int64
dtypes: datetime64[ns](1), int64(1), object(1)
memory usage: 248.0+ bytes