用一系列非数字对象中最接近的值替换 NaN？

Question

我正在使用 Pandas 和 Numpy，我正在尝试替换系列中的所有 NaN 值，如下所示：

date                    a
2017-04-24 01:00:00  [1,0,0]
2017-04-24 01:20:00  [1,0,0]
2017-04-24 01:40:00  NaN
2017-04-24 02:00:00  NaN
2017-04-24 02:20:00  [0,1,0]
2017-04-24 02:40:00  [1,0,0]
2017-04-24 03:00:00  NaN
2017-04-24 03:20:00  [0,0,1]
2017-04-24 03:40:00  NaN
2017-04-24 04:00:00  [1,0,0]

与最近的对象（在本例中为 Numpy 数组）。结果是：

date                    a
2017-04-24 01:00:00  [1,0,0]
2017-04-24 01:20:00  [1,0,0]
2017-04-24 01:40:00  [1,0,0]
2017-04-24 02:00:00  [0,1,0]
2017-04-24 02:20:00  [0,1,0]
2017-04-24 02:40:00  [1,0,0]
2017-04-24 03:00:00  [1,0,0]
2017-04-24 03:20:00  [0,0,1]
2017-04-24 03:40:00  [0,0,1]
2017-04-24 04:00:00  [1,0,0]

有人知道有效的方法吗？非常感谢。

Answer 1

删除空值然后用 reindex

填充

df.set_index('date').a.dropna().reindex(df.date, method='nearest').reset_index()

                 date          a
0 2017-04-24 01:00:00  [1, 0, 0]
1 2017-04-24 01:20:00  [1, 0, 0]
2 2017-04-24 01:40:00  [1, 0, 0]
3 2017-04-24 02:00:00  [0, 1, 0]
4 2017-04-24 02:20:00  [0, 1, 0]
5 2017-04-24 02:40:00  [1, 0, 0]
6 2017-04-24 03:00:00  [0, 0, 1]
7 2017-04-24 03:20:00  [0, 0, 1]
8 2017-04-24 03:40:00  [1, 0, 0]
9 2017-04-24 04:00:00  [1, 0, 0]

用一系列非数字对象中最接近的值替换 NaN？

Replace NaN with nearest value in a series of non-numeric object?

numpy

time-series

nan

python-2.7

pandas