Subgroup Discovery
A classic problem in data mining that aims to identify regions in feature space where the target variable deviates significantly from its global marginal distribution. Its core challenge is the trade-off between bias and support.
Timeline (last 90 days)
Subgroup discovery is a classic problem in data mining, aiming to find regions in the feature space where the target variable deviates from its marginal distribution.
Subgroup discovery has long faced a trade-off between deviation and support: small subgroups can deviate highly but lack statistical significance, while large subgroups are constrained by their construction and are hard to deviate substantially.
Existing subgroup discovery methods usually compress deviation and support into a single scalar quality measure, implicitly locking in one particular trade-off, which makes it difficult to recover the optimal subgroup under a user-specified support constraint.
A recent paper proposed the first theoretical and structural characterization of the maximum-KL-divergence subgroup under arbitrary support constraints.
In subgroup discovery, KL divergence is used to quantify the degree to which the target variable distribution within a subgroup deviates from the global marginal distribution; the larger the KL divergence, the more anomalous the subgroup.
All Facts (5)
Subgroup discovery is a classic problem in data mining, aiming to find regions in the feature space where the target variable deviates from its marginal distribution.
50%UnverifiedSubgroup discovery has long faced a trade-off between deviation and support: small subgroups can deviate highly but lack statistical significance, while large subgroups are constrained by their construction and are hard to deviate substantially.
50%UnverifiedExisting subgroup discovery methods usually compress deviation and support into a single scalar quality measure, implicitly locking in one particular trade-off, which makes it difficult to recover the optimal subgroup under a user-specified support constraint.
50%UnverifiedA recent paper proposed the first theoretical and structural characterization of the maximum-KL-divergence subgroup under arbitrary support constraints.
50%UnverifiedIn subgroup discovery, KL divergence is used to quantify the degree to which the target variable distribution within a subgroup deviates from the global marginal distribution; the larger the KL divergence, the more anomalous the subgroup.
50%