Classwise

Classwise is a term used across several disciplines to denote analysis, processing, or evaluation that is organized by class, category, or label. It functions as an adverbial or adjectival modifier to indicate that results or operations are considered on a per-class basis rather than across all data jointly.

In statistics and machine learning, classwise metrics are computed separately for each class in a multiclass setting. For example, classwise accuracy measures the proportion of correct predictions within each class, while classwise precision, recall, and F1 scores reflect performance specific to individual classes. These per-class metrics are often reported alongside aggregated measures such as macro- or micro-averaged scores to provide a more nuanced view of model behavior across different classes. The term highlights the dimension along which evaluation is performed rather than providing a single global statistic.

In software engineering and data processing, classwise processing refers to grouping data by class labels and applying operations to each group independently. This can include per-class normalization, validation, or transformation steps, enabling tailored handling for each category.

Etymology and usage notes: The suffix -wise in English conveys relation to a specified dimension. Classwise simply denotes consideration with respect to class or category. Because different fields adopt varying conventions, alternatives such as per-class or class-specific are common, and exact definitions can vary with context.

Related terms include per-class, macro-averaging, micro-averaging, and class-wise evaluation.