Mean absolute error (MAE) summarizes how far predictions are from actual values on average. It takes each prediction error without regard to direction, measures its absolute size, and averages those sizes. Smaller MAE indicates better predictive performance, but it does not make a model automatically useful: predictions can still vary widely around the regression line. section 3.4.2
Comparing shipping-cost models
For freight-value predictions, the product-weight model reported MAE=6.89, while the product-volume model reported MAE=7.41. The weight model therefore had the smaller average absolute error in these results. Interpret the values in the dataset’s freight-value units and compare them with MSE, RMSE, and R-squared rather than relying on MAE alone. section 4.6
MAE supports practical questions such as whether a model is accurate enough for shipping estimates. Validate the model with additional data and analysis, because its assumptions and the many other factors affecting real shipping costs can limit reliability. section 4.6
