Миноритарный класс — метка класса, которая встречается реже в несбалансированном classification dataset. Поскольку модели могут хуже работать на редких классах, производительность для minority class часто требует целевых метрик, resampling, weighting или специализированной оценки.
The less common label in a class-imbalanced dataset. For example, given a dataset containing 99% negative labels and 1% positive labels, the positive labels are the minority class. Contrast with majority class. A training set with a million examples sounds impressive. However, if the minority class is poorly represented, then even a very large training set might be insufficient. Focus less on the total number of examples in the dataset and more on the number of examples in the minority class. If your dataset doesn't contain enough minority class examples, consider using downsampling (the definition in the second bullet) to supplement the minority class. See Datasets: Imbalanced datasets in Machine Learning Crash Course for more information.