触及尾部：数据稀缺下校准多样性驱动保形覆盖率
Reaching the Tail: Calibration Diversity Drives Conformal Coverage under Data Scarcity
- University of Moratuwa（莫拉图瓦大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本文针对长宏观经济序列数据稀缺下的多期罕见事件预测难题，提出基于校准多样性的策略，将六个月覆盖率从67.8%提升至81.4%，并量化了诚实评分下六个月覆盖率能否达90%的待解问题。
AI中文摘要：
在长宏观经济序列的数据约束下，多期罕见事件预测难度较大：标注事件稀缺，而标准不确定性量化假设的可交换性与自相关性相违背。受控消融实验显示，自适应保形推理（Adaptive Conformal Inference）的表观罕见事件阈值实则反映校准集大小。在200个随机校准集中，非一致性得分分布的支持宽度可解释85%的覆盖率方差，而罕见事件计数仅解释2%；该结论在合成条件与五个国家中重复出现（五国斯皮尔曼相关系数ρ为0.45-0.66，而罕见事件计数的ρ为0.02-0.23）。基于此构建的多样性最大化选择器是所测试策略中唯一能提升长期覆盖率的方法（六个月时从67.8%提升至81.4%）；Mondrian、抗分布偏移及极值替代方法均未能缩小覆盖率差距，Mondrian甚至在使用最优标注时会恶化覆盖率。一个简洁命题解释了原因：覆盖率缺陷反映校准集的上侧分位数触及测试分布的紧密程度。多样性是必要非充分条件。该结论在含RegressorChain的两阶段美国衰退预测框架中得到验证，六个月覆盖率能否在诚实评分下达到90%仍未可知，本文量化了该问题而非解决它。
英文摘要：
Multi-horizon rare-event forecasting is hard under long macroeconomic series' data constraints: labeled events are scarce, and standard uncertainty quantification assumes an exchangeability that autocorrelation violates. A controlled ablation shows an apparent rare-event threshold for Adaptive Conformal Inference instead reflects calibration-set size. Across 200 random calibration sets, support width of the nonconformity-score distribution explains up to 85% of coverage variance versus 2% for rare-event count; the same, not the same magnitude, replicates across synthetic conditions and five countries (five-country Spearman $ρ$ 0.45-0.66 vs. 0.02-0.23). A diversity-maximizing selector built on this is the only strategy tested that improves long-horizon coverage (67.8% to 81.4% at six months); Mondrian, shift-robust, and extreme-value alternatives fail to close it. Mondrian even worsens coverage under oracle labels. A compact proposition explains why: coverage deficit reflects how closely the calibration set's upper quantile reaches the test distribution's. Diversity is necessary, not sufficient. Demonstrated on a two-stage U.S. recession-forecasting framework with RegressorChain, whether six-month coverage reaches 90% under honest scoring remains open, a question this paper quantifies rather than resolves.
