如何通过Python插值3d时间序列数据？

Question

I have some 3d time series data. But they have some missing values(np.nan). So, I want to interpolate them.

Example:

x = [0, nan, 2, 3, 4, 5]
y = [0, nan, 2, 3, 4, 5]
z = [3, nan, 5, 6, 7, 8]

# I want
interp(1) -> [1, 1, 4] 

How can I calculate?

I tried to interpolate x, y, and z for each individually by spline. But I feel it is strange. Is it strange to interpolate individually or is it right mathematically or is there any solutions?

In case calculating individually, it's very easy using scipy:

fx = interpolate.interp1d(time, x, kind="cubic")
fy = interpolate.interp1d(time, y, kind="cubic")
fz = interpolate.interp1d(time, z, kind="cubic")

Answer 1

Based on your comment it looks like you want to replace nan values with average of both sides (even though you've mentioned spline method). So here's a simple function in case nan does not happen at the start and end. import numpy as np

x = np.array([0, np.nan, 2, 3, 4, 5])
y = np.array([0, np.nan, 2, 3, 4, 5])
z = np.array([3, np.nan, 5, 6, 7, 8])
def nan_mean(v):
    ind=np.where(np.isnan(v))[0]
    v[ind]=(v[ind+1]+v[ind-1])/2
nan_mean(x)
nan_mean(y)
nan_mean(z)

this returns your desired value