base rate

The base rate is the underlying frequency or proability of a phenomena. For example, if one in thousand people are likely to be suffering from flu during the winter, then the base rate of flu is 0.001. The base rate is critical when using evidence to make diagnoses or other decisions. For example, if flu and the common cold have similar symptoms and the base rate of the common cold is one in 50, then someone with a runny nose and headache is more likely to be suffering from a cold than flu. This is captured formally in Bayes Theorem.

Base rates are also important when trying to reduce bias or discrimination as we try to be fair to an individual irrespectve of the base rate for their demographic. Without taking remedial action, data-based machine learning can be discriminatory because of bate rate differences, even if the data is free from human bias.