Conditional Decomposition Discrepancy: Separating Conditional Bias from Sampling Bias in Small-Sample Cross-Domain Cost Forecasting
- 1Graduate School, Air Force Engineering University, Xi’an 710051, China
- 2Aviation Engineering School, Air Force Engineering University, Xi’an 710038, China
- 3Aviation Maintenance NCO Academy, Air Force Engineering University, Xinyang 464000, China
Abstract
Sample discrepancy measurement is an essential step in transfer learning models such as cross-domain cost forecasting. However, tools that can effectively separate conditional bias from sampling bias are still lacking, so discrepancy information cannot fully support sample screening and model-update decisions. This paper analyzes the information boundary of the maximum mean discrepancy (MMD) in bias attribution, states and proves the relevant propositions, and proposes shifting the object of testing from the joint distribution to the conditional distribution. On this basis, the conditional decomposition discrepancy (CDD) measure is designed: a conditional mean and a residual bandwidth are estimated from an independent calibration sample, the data are projected onto a common residual axis and subjected to a kernel two-sample test, and the test result is converted into a continuous risk score with reference to the permutation null distribution. Numerical experiments are then designed on a UK property transaction price dataset. The results show that, provided the conditional difference is transmitted to the residual marginal, the proposed measure effectively detects conditional bias and reduces the proportion of contaminated source-domain samples under controlled injection, thus offering a new discrepancy measurement tool for data screening in transfer learning.
Related
- Conditional Kernel Stein Discrepancy
- Realized Tail-Risk Conditional Autoregressive Expectile Models: Accurate and Stable Forecasting Performance
- Online Conditional Vine Copulas: Forecasting Electricity Demand
- Regime-Aware Estimation of Large Equity Correlation Matrices: Tail-Pairwise Dependence and Dynamic Conditional Correlation Decomposition for Water and Energy Sectors
- A unifying theory for the evaluation of conditional Akaike information for mixed-effects models
BibTeX
@article{li2026conditional,
title = {{Conditional Decomposition Discrepancy: Separating Conditional Bias from Sampling Bias in Small-Sample Cross-Domain Cost Forecasting}},
author = {Bo Li and Jilian Guo and Yingjie Luo and Xiaofei Chi},
journal = {Mathematics},
year = {2026},
month = nov,
doi = {10.3390/math14213986},
url = {https://www.mdpi.com/2227-7390/14/21/3986},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Conditional Decomposition Discrepancy: Separating Conditional Bias from Sampling Bias in Small-Sample Cross-Domain Cost Forecasting
AU - Bo Li
AU - Jilian Guo
AU - Yingjie Luo
AU - Xiaofei Chi
PY - 2026
DA - 2026/11/02
JO - Mathematics
DO - 10.3390/math14213986
UR - https://www.mdpi.com/2227-7390/14/21/3986
AB - Sample discrepancy measurement is an essential step in transfer learning models such as cross-domain cost forecasting. However, tools that can effectively separate conditional bias from sampling bias are still lacking, so discrepancy information cannot fully support sample screening and model-update decisions. This paper analyzes the information boundary of the maximum mean discrepancy (MMD) in bias attribution, states and proves the relevant propositions, and proposes shifting the object of testing from the joint distribution to the conditional distribution. On this basis, the conditional decomposition discrepancy (CDD) measure is designed: a conditional mean and a residual bandwidth are estimated from an independent calibration sample, the data are projected onto a common residual axis and subjected to a kernel two-sample test, and the test result is converted into a continuous risk score with reference to the permutation null distribution. Numerical experiments are then designed on a UK property transaction price dataset. The results show that, provided the conditional difference is transmitted to the residual marginal, the proposed measure effectively detects conditional bias and reduces the proportion of contaminated source-domain samples under controlled injection, thus offering a new discrepancy measurement tool for data screening in transfer learning.
SN - 2227-7390
ER -