Python中带x值的缺失数据插值工具选择与实现方法咨询

If you're dealing with missing y-values paired with known x-values, two reliable tools in Python are scipy.interpolate.interp1d and pandas.Series.interpolate. Let's break down how to implement each, including fixes for common pitfalls:

1. Using scipy.interpolate.interp1d #

First heads-up: interp1d can't handle NaN values directly—you'll need to filter out the missing data points first. Here's a working example:

import numpy as np
import matplotlib.pyplot as plt
from scipy import interpolate

# Create data with missing y-values
x = np.arange(0, 10)
y = x**2 + x**3
y[4] = np.nan
y[7] = np.nan

# Filter out pairs where y is missing
valid_mask = ~np.isnan(y)
x_valid = x[valid_mask]
y_valid = y[valid_mask]

# Create interpolation function (default is linear; use 'quadratic'/'cubic' for smoother results)
f = interpolate.interp1d(x_valid, y_valid, kind='linear')

# Generate interpolated values for all original x points
ynew = f(x)

# Optional: Visualize to check results
plt.plot(x, y, 'o', label='Original (with NaNs)')
plt.plot(x, ynew, '-', label='Interpolated')
plt.legend()
plt.show()

Key takeaways:

• Use the kind parameter to pick your interpolation style (linear is safest for most cases, cubic works well for smooth data)
• This method requires your x-values to be sorted (which is typical for sequential data like time series or ordered measurements)

2. Using pandas.Series.interpolate #

Pandas makes this way simpler—it automatically handles NaN values without manual filtering, and plays nicely with labeled data. Here's how to use it:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# Create data with missing y-values
x = np.arange(0, 10)
y = x**2 + x**3
y[4] = np.nan
y[7] = np.nan

# Convert to a Series, using x as the index to align values properly
yp = pd.Series(y, index=x)

# Run interpolation (default is linear; specify method for different behaviors)
yp_interpolated = yp.interpolate(method='linear')

# Optional: Plot to verify
plt.plot(x, y, 'o', label='Original (with NaNs)')
plt.plot(x, yp_interpolated, '-', label='Pandas Interpolated')
plt.legend()
plt.show()

Key takeaways:

• Popular method options: 'linear', 'quadratic', 'cubic', 'nearest'
• Using x as the Series index ensures interpolation uses the actual x-values, not just the position of the data point (critical if your x isn't sequential)
• For time-series data, pandas also supports time-aware interpolation with method='time'

Which tool should you pick? Use scipy if you need to reuse the interpolation function for new x-values later, and pandas for quick, clean handling of tabular or series data with minimal code.

内容的提问来源于stack exchange，提问作者Ammar