Null Accuracy
Null Accuracy is a baseline classification metric that represents the accuracy a model would achieve by always predicting the most common class in a dataset, without using the input features.
What is Null Accuracy?
Null Accuracy is calculated by identifying the class with the highest number of examples and dividing its frequency by the total number of examples. For example, if 80% of a dataset belongs to one class, the Null Accuracy is 80%. A trained classifier should generally perform better than this baseline to demonstrate useful predictive capability.
Why is Null Accuracy Important?
Null Accuracy provides a simple benchmark for evaluating classification models. It is particularly useful for identifying class imbalance, where a model can appear highly accurate simply by predicting the majority class while performing poorly on less frequent classes.
Common use cases
Null Accuracy is commonly used as a baseline for binary and multiclass classification, model evaluation, class imbalance analysis, and comparing classifier performance.