When estimating shipping costs, mean squared error (MSE) measures how far predictions are from actual values by averaging squared differences. Squaring makes large mistakes count more heavily, so a lower MSE indicates better predictive performance. section 3.4.2
MSE is useful alongside other regression metrics rather than as a verdict by itself. Mean absolute error (MAE) is easier to read as an average error, while root mean squared error (RMSE) returns to the target’s units. Residual analysis and diagnostic plots help examine model assumptions. section 3.4.2
A practical workflow uses MSE to compare shipping-cost models:
- Compare MSE with MAE, RMSE, and R2 instead of treating one score as conclusive. section 4.6
Comparing weight and volume models
For the shipping data, the weight model reports MSE=156.77, while the volume model reports MSE=163.70. The weight model performs slightly better on this measure, but both models explain only part of shipping-cost variation: their R2 values are 0.37 and 0.34. Validate such conclusions with further data and analysis. section 4.6
