问题描述
我想根据DatetimeIndex条件将一个数据帧中的值分配给另一个数据帧。
date importance
2006-12-05 10:35:00 HIGH
2006-12-13 02:40:00 LOW
此数据框：（第二个）
index value
2006-12-05 08:03:01.985 6
2006-12-05 08:11:34.130 7
2006-12-05 08:20:05.959 6
2006-12-05 08:28:38.104 6
2006-12-05 08:37:02.995 6
2006-12-05 08:45:35.140 5
2006-12-05 08:54:06.969 6
2006-12-05 09:02:59.928 6
2006-12-05 09:11:32.072 6
2006-12-05 09:20:03.901 6
2006-12-05 09:28:36.046 5
2006-12-05 09:37:00.937 5
2006-12-05 09:45:33.082 6
2006-12-05 09:54:04.911 6
2006-12-05 10:02:04.889 6
2006-12-05 10:10:37.034 5
2006-12-05 10:19:08.863 6
2006-12-05 10:27:41.008 5
2006-12-05 10:36:04.953 5
2006-12-05 10:44:37.098 5
.
.
.
2006-12-13 02:06:00.898 1
2006-12-13 02:14:33.043 1
2006-12-13 02:23:04.872 1
2006-12-13 02:31:03.904 1
2006-12-13 02:39:36.048 1
2006-12-13 02:48:07.878 2
2006-12-13 02:56:40.022 5
2006-12-13 03:05:04.914 2
2006-12-13 03:13:37.058 3
2006-12-13 03:22:08.888 6
2006-12-13 03:31:03.108 1
2006-12-13 03:39:34.937 1
2006-12-13 03:48:07.081 1
2006-12-13 03:56:38.911 2
2006-12-13 04:05:04.117 3
最终结果应该是这样：
index value new_value
2006-12-05 08:03:01.985 6
2006-12-05 08:11:34.130 7
2006-12-05 08:20:05.959 6
2006-12-05 08:28:38.104 6
2006-12-05 08:37:02.995 6
2006-12-05 08:45:35.140 5
2006-12-05 08:54:06.969 6
2006-12-05 09:02:59.928 6
2006-12-05 09:11:32.072 6
2006-12-05 09:20:03.901 6
2006-12-05 09:28:36.046 5
2006-12-05 09:37:00.937 5
2006-12-05 09:45:33.082 6
2006-12-05 09:54:04.911 6
2006-12-05 10:02:04.889 6
2006-12-05 10:10:37.034 5
2006-12-05 10:19:08.863 6
2006-12-05 10:27:41.008 5
2006-12-05 10:36:04.953 5 HIGH
2006-12-05 10:44:37.098 5
.
.
.
2006-12-13 02:06:00.898 1
2006-12-13 02:14:33.043 1
2006-12-13 02:23:04.872 1
2006-12-13 02:31:03.904 1
2006-12-13 02:39:36.048 1 LOW
2006-12-13 02:48:07.878 2
2006-12-13 02:56:40.022 5
2006-12-13 03:05:04.914 2
2006-12-13 03:13:37.058 3
2006-12-13 03:22:08.888 6
2006-12-13 03:31:03.108 1
2006-12-13 03:39:34.937 1
2006-12-13 03:48:07.081 1
2006-12-13 03:56:38.911 2
2006-12-13 04:05:04.117 3
我尝试过：
def getNearestDate(items,pivot):
return min(items,key=lambda x: abs(x - pivot))
items = second_df.index
for pivot in first_df.date:
d = getNearestDate(items,pivot)
print(d)
second_df.loc[second_df.index == d,'new_value'] = first_df.importance
它会打印以下最接近的日期：
2006-12-05 10:36:04.953000
2006-12-13 02:39:36.048000
因此，这几天应该将“重要性”中的值放入。
另外，在new_value列上，所有内容均为NAN。
您能帮我解决这个问题吗？
解决方法
您已使用loc中的条件
second_df.index == d
并在满足条件的索引处返回true，而不是索引。
代替使用
second_df[second_df.index == d].index.values
您已经有了for (int i = 1; i < n; ++i) {
if (arr[i - 1] == 0) {
insert at i a 0;
}
}
insert at i a 0:
// First move the remaining to the right: i .. n-2
...
// Then fill in the zero
arr[i] = 0;
所需的遮罩。这样会生成一个pandas.Series，其值second_df.index == d为真，而True为假。您可以False一起使用多个掩码，以获取任何掩码中|=的所有行。只需将该系列作为“ new_value”列添加到第二个数据框即可。
True如果您确实希望mask = False
for pivot in first_df.date:
mask |= second_df.index == getNearestDate(second_df.index,pivot)
second_df['new_value'] = mask
和'X'为''和True的别名，则还可以在添加系列到数据框。
False编辑：
如果要获取第一个数据帧的mask = False
for pivot in first_df.date:
mask |= second_df.index == getNearestDate(second_df.index,pivot)
second_df['new_value'] = mask.apply(lambda x: 'X' if bool(x) else '')
值，则可以简单地使用getNearestDate函数来确定哪些行需要该值，然后将它们与第二个数据帧合并。
importance您应该只能使用reindex和merge
# note the method and the tolerance. Change them to whatever works best for your actual data
new_df = df2.merge(df.reindex(df2.index,method='nearest',limit=1,tolerance='2T'),left_index=True,right_index=True)
value importance
index
2006-12-05 08:03:01.985 6 NaN
2006-12-05 08:11:34.130 7 NaN
2006-12-05 08:20:05.959 6 NaN
2006-12-05 08:28:38.104 6 NaN
2006-12-05 08:37:02.995 6 NaN
2006-12-05 08:45:35.140 5 NaN
2006-12-05 08:54:06.969 6 NaN
2006-12-05 09:02:59.928 6 NaN
2006-12-05 09:11:32.072 6 NaN
2006-12-05 09:20:03.901 6 NaN
2006-12-05 09:28:36.046 5 NaN
2006-12-05 09:37:00.937 5 NaN
2006-12-05 09:45:33.082 6 NaN
2006-12-05 09:54:04.911 6 NaN
2006-12-05 10:02:04.889 6 NaN
2006-12-05 10:10:37.034 5 NaN
2006-12-05 10:19:08.863 6 NaN
2006-12-05 10:27:41.008 5 NaN
2006-12-05 10:36:04.953 5 HIGH
2006-12-05 10:44:37.098 5 NaN
2006-12-13 02:06:00.898 1 NaN
2006-12-13 02:14:33.043 1 NaN
2006-12-13 02:23:04.872 1 NaN
2006-12-13 02:31:03.904 1 NaN
2006-12-13 02:39:36.048 1 LOW
2006-12-13 02:48:07.878 2 NaN
2006-12-13 02:56:40.022 5 NaN
2006-12-13 03:05:04.914 2 NaN
2006-12-13 03:13:37.058 3 NaN
2006-12-13 03:22:08.888 6 NaN
2006-12-13 03:31:03.108 1 NaN
2006-12-13 03:39:34.937 1 NaN
2006-12-13 03:48:07.081 1 NaN
2006-12-13 03:56:38.911 2 NaN
2006-12-13 04:05:04.117 3 NaN
只需进行这些小的更改，希望它就会起作用
loc=[]
def getNearestDate(items,pivot):
return min(items,key=lambda x: abs(x - pivot))
items = second_df.index
for pivot in first_df.date:
d = getNearestDate(items,pivot)
loc.append(second_df.set_index('index').index.get_loc(d))
## Adding Data to your second df
second_df['importance']=[]
for index,locations in enumerate(loc):
df['importance'][int(location)]=first_df['importance'][index]
首先，我们必须保留与原始数据框中的日期相对应的日期：
items = second_df.index
dates = []
for pivot in first_df.date:
dates.append(getNearestDate(items,pivot))
first_df['new_date'] = dates
由于我们不再需要它们，因此可以删除整列：
first_df = first_df.drop(columns="date")
为了使合并生效，我们需要在两个数据帧上都声明索引。
first_df.set_index("new_date",inplace =True)
合并过程如下：
second_df = second_df.merge(first_df,how = "left",right_index=True)
此外，重要的是永远不要让NaN出现在数据框中：
second_df.importance = second_df.importance.fillna(0)