Accuracy

Accuracy provides a straightforward measure of model performance by calculating the ratio of correct outcomes to total outcomes. It answers the question "How often did the model make the right decision?". The metric treats all correct predictions correctly, regardless of class.

Calculation

The accuracy is calculated as:

\(\frac{\text{true positives} + \text{true negatives}}{\text{true positives} + \text{true negatives} + \text{false positives} + \text{false negatives}}\)

Key Characteristics

Accuracy ranges from 0 to 1, with 1 representing the best possible accuracy.

Limitations

Accuracy can be misleading, especially for cases with imbalanced datasets. Consider for example, a patient health prediction model. For a class distribution with 95% healthy patients, a model which always predicts healthy will achieve an accuracy of 95%.