Unverified50% confidenceFactExact time
分层回测将股票按因子值分为若干组（通常5组或10组），观察各组超额收益是否呈现单调分布，比单一IC指标更直观揭示因子非线性特征
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
UnverifiedRank IC改用Spearman秩相关，只关心排序关系，对极端值更鲁棒，是A股研究中更常用的指标64% similarUnverifiedThe --sref weight value has a maximum of 10; in this case it is set to 7, and the higher the value, the closer the generated result is to the reference style.63% similarUnverifiedIn highly imbalanced datasets (with a positive rate below 5%), the area under the precision-recall curve (AP) often reflects the model's ability to detect the minority class better than AUC does.61% similarUnverifiedAssuming the data follows a Gaussian distribution, maximizing the likelihood is equivalent to minimizing the mean squared error.59% similarUnverified综合多个独立来源的评测结果比单一自报基准分数更能反映模型的实际水平59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/584899API
curl https://kongchang.com/api/v1/knowledge/claims/584899MCP
get_claim(id=584899)