如何将NumPy数组形式的SARIMA预测结果添加到Pandas DataFrame？

Hey there! Let's work through this issue step by step to get your SARIMAX predictions properly integrated into your DataFrame, then handle the plotting and metric calculations.

1. Root Cause: Dimension Mismatch #

The problem you're seeing almost always comes down to mismatched dimensions between your predictions array and the test DataFrame, or a 2D prediction array being treated as a single nested list by Pandas.

First, let's diagnose the shape of your data:

# Check the shape of your predictions and test DataFrame
print("Predictions shape:", predictions.shape)
print("Test DataFrame rows:", test.shape[0])

If predictions shows a 2D shape (like (1, 100) instead of (100,)), flatten it to a 1D array first:

predictions = predictions.flatten()

Then you can directly assign it to the DataFrame—no need to convert to a list (Pandas plays nicely with NumPy arrays):

test['predicted'] = predictions

Double-check your SARIMAX prediction parameters too! Make sure you're predicting exactly the length of your test set, e.g.:

# Example: Predict from the end of training data to end of test data
predictions = model.predict(start=len(train), end=len(train)+len(test)-1)

2. Verify the New Column #

After assignment, confirm everything looks right with:

print(test[['your_actual_value_column', 'predicted']].head())

You should see a separate prediction value for every row in the test set.

3. Plot Actual vs. Predicted Values #

Use Matplotlib for a clean comparison plot—replace your_actual_value_column with your test set's true value column name:

import matplotlib.pyplot as plt

plt.figure(figsize=(12, 6))
plt.plot(test['your_actual_value_column'], label='Actual Values', linewidth=2)
plt.plot(test['predicted'], label='Predicted Values', linestyle='--', linewidth=2)
plt.title('Time Series: Actual vs. SARIMAX Predictions')
plt.xlabel('Date/Time')
plt.ylabel('Value')
plt.legend()
plt.grid(alpha=0.3)
plt.show()

4. Calculate MSE and RMSE Metrics #

You can use scikit-learn's metrics module for this, or compute manually if you prefer:

Option 1: Using scikit-learn #

from sklearn.metrics import mean_squared_error
import numpy as np

mse = mean_squared_error(test['your_actual_value_column'], test['predicted'])
rmse = np.sqrt(mse)

print(f"Mean Squared Error (MSE): {mse:.4f}")
print(f"Root Mean Squared Error (RMSE): {rmse:.4f}")

Option 2: Manual Calculation #

import numpy as np

mse = ((test['your_actual_value_column'] - test['predicted']) ** 2).mean()
rmse = np.sqrt(mse)

print(f"MSE: {mse:.4f} | RMSE: {rmse:.4f}")

Quick Troubleshooting Checks #

• Ensure your SARIMAX prediction window exactly matches the time range of your test set
• Confirm predictions is a 1D array (not nested) before assigning to the DataFrame
• Make sure your test DataFrame has no missing rows that would throw off the length match

内容的提问来源于stack exchange，提问作者bbartling