Question

我正在尝试在时间序列数据之间重新采样和内插。由于某些原因，我不断获得NA的行。

import pandas as pd

track = pd.DataFrame({'latitude': [32.743680275092686, 32.75168725999735,
                                   32.7782220355535, 32.783715199616,
                                   32.78683419954979, 32.80517578125,
                                   32.809039616988876, 32.82146906448623,
                                   32.83804166114936, 32.839903750662074],
                     'longitude': [-111.19133774115113, -111.205463020169,
                                   -111.25232307278381, -111.26207624162947,
                                   -111.26762545838648, -111.30013602120536,
                                   -111.30697444993622, -111.32877894810267,
                                   -111.35798240194515, -111.36134556361606],
                     'altitude':  [21125.0, 20975.0,
                                   20175.0, 20000.0,
                                   19900.0, 19350.0,
                                   19225.0, 18850.0,
                                   18325.0, 18250.0]},
                  index=[pd.Timestamp('2016-10-20 20:57:24.189484'),
                         pd.Timestamp('2016-10-20 20:57:32.102171'),
                         pd.Timestamp('2016-10-20 20:57:58.872157'),
                         pd.Timestamp('2016-10-20 20:58:04.522744'),
                         pd.Timestamp('2016-10-20 20:58:07.806292'),
                         pd.Timestamp('2016-10-20 20:58:26.574050'),
                         pd.Timestamp('2016-10-20 20:58:30.495372'),
                         pd.Timestamp('2016-10-20 20:58:43.221657'),
                         pd.Timestamp('2016-10-20 20:59:00.496501'),
                         pd.Timestamp('2016-10-20 20:59:02.362993')])

track.resample('S').interpolate('linear')

重新采样和内插的结果完全用NaN填充：

                     latitude  longitude  altitude
2016-10-20 20:57:24       NaN        NaN       NaN
2016-10-20 20:57:25       NaN        NaN       NaN
2016-10-20 20:57:26       NaN        NaN       NaN
2016-10-20 20:57:27       NaN        NaN       NaN
2016-10-20 20:57:28       NaN        NaN       NaN
...

我怀疑想要获取具有毫秒分辨率的点并在它们之间的1秒范围内进行插值会很复杂。我可以将索引舍入到最接近的秒数，并使结果更接近我的期望：

track.index = track.index.round('1s')
track.resample('S').interpolate('linear')

结果：

                      latitude   longitude      altitude
2016-10-20 20:57:24  32.743680 -111.191338  21125.000000
2016-10-20 20:57:25  32.744681 -111.193103  21106.250000
2016-10-20 20:57:26  32.745682 -111.194869  21087.500000
2016-10-20 20:57:27  32.746683 -111.196635  21068.750000
2016-10-20 20:57:28  32.747684 -111.198400  21050.000000
...

舍入会引入一定程度的误差，尽管在这种情况下很小。有更好的方法吗？

熊猫根据重新采样的NA进行插值

0 个答案: