characteristichave

Characteristichave is a theoretical construct in data modeling that describes a minimal and sufficient set of attributes that uniquely identifies an object within a given domain.

Formally, for a domain U and a set of attributes F, a subset H ⊆ F is a characteristichave for an object x ∈ U if the projection of x on H is unique in U, and no proper subset of H is likewise unique.

Characteristichaves are closely related to candidate keys in relational databases and to discriminative feature sets in machine learning. They are typically context-dependent and can vary with data distribution, sampling, and preprocessing.

Applications include schema design, record linkage, data deduplication, and privacy risk assessment. Because a characteristichave can reveal identifying combinations of attributes, systems may minimize or generalize these attributes to reduce reidentification risk.

Example: in a bibliographic catalog, a practical characteristichave might be the combination of ISBN and edition, which together distinguish a specific published work.