Yes, the innocent do perish.
 |
| Rendy Novantino |
DATA SCIENTISTS WORRY about underfitting, a phenomenon in which a formula predicts outcomes using limited inputs.
As you can imagine, this results in output errors because the underfit model fails to account for a wide enough variety of possibilities.
Underfitting, then, is essentially an oversimplification of a causal relationship based on insufficient data points. One example, spelled out by the AI wizards at DataRobot, is the correlation between sales and advertising:
An underfitted model may suggest that you can always make better sales by spending more on marketing when in fact the model fails to capture a saturation effect (at some point, sales will flatten out no matter how much more you spend on marketing).
So while it's generally true that spending more money on advertising will yield increased sales, at some point marketers will see diminishing returns on investment. The underfit model might suggest a straight line when, in reality, the model should look something more like a logarithmic function, curving upwards until flattening out at a certain point.
In a similar way we see Eliphaz, in the fourth chapter of Job, underfitting one of his proverbs; he attempts to derive a universal truth from a generally true axiom.
