Accuracy vs F1 score

Why they differ

Accuracy counts every correct answer alike, so it hides a rare class that is always missed; the F1 score looks only at that class and exposes the problem.

Accuracy

Evaluation & metrics

The share of all cases a model gets right; simple to read, but it can look excellent when the thing you care about is rare.

Formal

In classification, the number of correct answers divided by the total number of cases; in a confusion matrix it is the sum of the cells where the answer and the true class agree, divided by the sum of all cells.

In plain English

Like a weather forecaster in the desert who says “no rain” every single day and is right almost every time, yet useless on the one day it pours.

In practice

A data analyst at a shipping company tests a model meant to warn of engine faults; it scores 99.5% accuracy, yet it has never raised a single warning, because faults occur on only one voyage in 200.

Why it matters

It is the number most people quote first, so teams must check whether the classes are balanced before trusting it, and turn to precision, recall or the F1 score when they are not.

F1 score

Evaluation & metrics

One number that blends precision and recall, so it stays low unless a model both finds the real cases and avoids wrong flags.

Formal

Two times precision times recall, divided by precision plus recall, a balanced average that is pulled down by whichever of the two is lower, computed for one class.

In plain English

Like judging a goalkeeper on both saves and clean kicks, where being perfect at one and hopeless at the other still earns a poor mark.

In practice

The IT operations lead at an accounting firm compares two spam filters, one with 95% precision and 40% recall, the other with 80% and 78%; their F1 scores, about 56% against 79%, make the second the clear choice.

Why it matters

It gives one fair number for comparing models on rare classes, where accuracy misleads, though it hides which of the two errors is growing.

Shared connections

Model evaluation (evals)