Comprehensive Causality Multi-Objective Counterfactual Explanation with Feature Selection

LIU Jin-ping; TANG Hao-nan; LI Xing-wang; XU Peng-fei; YUAN Sheng-wei

doi:10.12263/DZXB.20241166

您当前的位置：

首页 >

文章列表页 >

Comprehensive Causality Multi-Objective Counterfactual Explanation with Feature Selection

Large-Scale Models and the Internet | 更新时间：2025-10-16

- Comprehensive Causality Multi-Objective Counterfactual Explanation with Feature Selection
- ACTA ELECTRONICA SINICA Vol. 53, Issue 6, Pages: 1805-1814(2025)
- 作者机构：
  
  湖南师范大学信息科学与工程学院，湖南长沙 410081
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China(62371187)
- DOI：10.12263/DZXB.20241166
  CLC： TP181;
- Received：27 December 2024，
  
  Revised：2025-06-14，
  
  Published：25 June 2025
- 稿件说明：
移动端阅览
刘金平, 汤浩楠, 李兴旺, 等. 带特征选择的综合因果多目标反事实解释方法[J]. 电子学报, 2025, 53(06): 1805-1814.

LIU Jin-ping, TANG Hao-nan, LI Xing-wang, et al. Comprehensive Causality Multi-Objective Counterfactual Explanation with Feature Selection[J]. Acta Electronica Sinica, 2025, 53(06): 1805-1814.
刘金平, 汤浩楠, 李兴旺, 等. 带特征选择的综合因果多目标反事实解释方法[J]. 电子学报, 2025, 53(06): 1805-1814. DOI：10.12263/DZXB.20241166

LIU Jin-ping, TANG Hao-nan, LI Xing-wang, et al. Comprehensive Causality Multi-Objective Counterfactual Explanation with Feature Selection[J]. Acta Electronica Sinica, 2025, 53(06): 1805-1814. DOI：10.12263/DZXB.20241166

摘要

随着复杂机器学习模型应用扩展，各行业对模型可解释性的需求剧增.反事实解释是重要的事后可解释方法，但传统方法常将多目标合并为单目标优化，导致权重分配困难且难以调和目标冲突，也因忽略因果关系使生成的反事实样本不现实.此外，现有方法在高维、冗余、噪声数据下存在计算效率低、预测精度下降及全局解释不足等问题.为此，本文提出综合因果多目标反事实解释方法（Comprehensive Causal multi-objective counterfactual Explanation with Feature Selection，CCE-FS）.该方法首先基于最大互信息系数筛选关键特征以提升预测精度和全局解释力，然后将反事实搜索转化为多目标优化问题，有效平衡多目标关系.同时引入领域因果关系约束，确保反事实样本现实合理.CCE-FS还提供可视化特征效应分析，增强用户理解并揭示模型偏见.Statlog数据集实验表明，CCE-FS通过特征选择显著提高了反事实样本的有效性、正常性、稀疏度，并使连续特征接近度提升46.3%.在Adult-Income和COMPAS数据集上的验证进一步证明，CCE-FS在因果一致性、数据分布合理性和连续特征邻近度方面均优于现有方法，展现了更强的解释与应用潜力.

Abstract

The widespread adoption of complex machine learning models across diverse industries has significantly increased the demand for model interpretability. The counterfactual explanation is a crucial post-hoc explanation method. However

traditional approaches often combine multiple objectives into a single objective optimization problem

leading to difficulties in weight assignment and reconciling conflicting objectives. Furthermore

existing methods also suffer from low computational efficiency

degraded prediction accuracy

and insufficient global explanations when dealing with high-dimensional

redundant

and noisy data. To address these issues

this article proposes a comprehensive causal multi-objective counterfactual explanation method with feature selection (CCE-FS). CCE-FS first employs the maximal information coefficient (MIC) to select key features

thereby enhancing prediction accuracy and global explanatory power. It then formulates the counterfactual search as a multi-objective optimization problem

effectively balancing the relationships between multiple objectives. Domain-specific causal relationships are incorporated as constraints to ensure the generated counterfactuals are realistic and plausible. Additionally

CCE-FS provides visual feature effect analysis to enhance user understanding and reveal potential model biases. Experiments conducted on the Statlog dataset demonstrate that CCE-FS significantly improves the validity

normality

and sparsity of counterfactual samples through feature selection

achieving a 46.3% enhancement in proximity for continuous features. Further validation on the Adult-Income and COMPAS datasets confirms that CCE-FS outperforms existing methods in causal consistency

data distribution reasonableness

and proximity of continuous features. These results highlight CCE-FS’s superior explanatory capabilities and greater application potential.

关键词

Keywords

references

蔡美玲 , 罗迪 , 肖敬日 , 等 . 连续与离散变量协同分析的非平稳非高斯工业过程异常检测 [J ] . 电子学报 , 2024 , 52 ( 10 ): 3291 - 3300 .

CAI M L , LUO D , XIAO J R , et al . Continuous and discrete variables-concurrent analysis-based nonstationary and non-Gaussian industrial process anomaly detection [J ] . Acta Electronica Sinica , 2024 , 52 ( 10 ): 3291 - 3300 . (in Chinese)

刘金平 , 吴娟娟 , 张荣 , 等 . 基于结构重参数化与多尺度深度监督的COVID-19胸部CT图像自动分割 [J ] . 电子学报 , 2023 , 51 ( 5 ): 1163 - 1171 .

LIU J P , WU J J , ZHANG R , et al . Toward automated segmentation of COVID-19 chest CT images based on structural reparameterization and multi-scale deep supervision [J ] . Acta Electronica Sinica , 2023 , 51 ( 5 ): 1163 - 1171 . (in Chinese)

苏越阳 , 姚迪 , 毕经平 . 基于噪声标签重加权的车辆轨迹异常检测方法 [J ] . 电子学报 , 2025 , 53 ( 1 ): 182 - 192 .

SU Y Y , YAO D , BI J P . A vehicle trajectory anomaly detection method based on noise label re-weighting [J ] . Acta Electronica Sinica , 2025 , 53 ( 1 ): 182 - 192 . (in Chinese)

NAZEMI A , FABOZZI F J . Interpretable machine learning for creditor recovery rates [J ] . Journal of Banking Finance , 2024 , 164 : 107187 .

CHEN V , YANG M Y , CUI W B , et al . Applying interpretable machine learning in computational biology-pitfalls, recommendations and opportunities for new developments [J ] . Nature Methods , 2024 , 21 ( 8 ): 1454 - 1461 .

KARIMI A H , BARTHE G , SCHÖLKOPF B , et al . A survey of algorithmic recourse: Contrastive explanations and consequential recommendations [J ] . ACM Computing Surveys , 2023 , 55 ( 5 ): 1 - 29 .

WACHTER S , MITTELSTADT B , RUSSELL C . Counterfactual explanations without opening the black box: Automated decisions and the GDPR [J ] . Harvard Journal Of Law Technology , 2017 , 31 : 841 .

AUGUSTIN M , BOREIKO V , CROCE F , et al . Diffusion visual counterfactual explanations [J ] . Advances In Neural Information Processing Systems , 2022 , 35 : 364 - 377 .

KENNY E M , KEANE M T . On generating plausible counterfactual and semi-factual explanations for deep learning [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2021 , 35 ( 13 ): 11575 - 11585 .

MOTHILAL R K , SHARMA A , TAN C H . Explaining machine learning classifiers through diverse counterfactual explanations [C ] // Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency . New York : ACM , 2020 : 607 - 617 .

POYIADZI R , SOKOL K , SANTOS-RODRIGUEZ R , et al . FACE: Feasible and actionable counterfactual explanations [C ] // Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society . New York : ACM , 2020 : 344 - 350 .

向许 , 于洪 , 张晓霞 , 等 . IsomapVSG-LIME: 一种新的模型无关解释方法 [J ] . 智能系统学报 , 2023 , 18 ( 4 ): 841 - 848 .

XIANG X , YU H , ZHANG X X , et al . IsomapVSG-LIME: A novel local interpretable model-agnostic explanations [J ] . CAAI Transactions on Intelligent Systems , 2023 , 18 ( 4 ): 841 - 848 . (in Chinese)

KANAMORI K , TAKAGI T , KOBAYASHI K , et al . DACE: Distribution-aware counterfactual explanation by mixed-integer linear optimization [C ] // Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence . Freiburg : IJCAI , 2020 : 2855 - 2862 .

CHO S H , SHIN K S . Feature-weighted counterfactual-based explanation for bankruptcy prediction [J ] . Expert Systems with Applications , 2023 , 216 : 119390 .

CHEN L S , FERNANDO H , YING Y M , et al . Three-way trade-off in multi-objective learning: Optimization, generalization and conflict-avoidance [J ] . Advances In Neural Information Processing Systems , 2024 , 36 : 1 - 53 .

DUONG T D , LI Q , XU G D . Achieving counterfactual fairness with imperfect structural causal model [J ] . Expert Systems with Applications , 2024 , 240 : 122411 .

KUMAR I E , VENKATASUBRAMANIAN S , SCHEIDEGGER C , et al . Problems with Shapley-value-based explanations as feature importance measures [EB/OL ] . ( 2022-06-30 )[ 2025-06-30 ] . https://arxiv.org/abs/2002.11097v2 https://arxiv.org/abs/2002.11097v2 .

WANG H J , LIANG Q X , HANCOCK J T , et al . Feature selection strategies: A comparative analysis of SHAP-value and importance-based methods [J ] . Journal of Big Data , 2024 , 11 ( 1 ): 44 .

MAKAROVA A , SHEN H , PERRONE V , et al . Overfitting in Bayesian optimization: An empirical study and early-stopping solution [C ] // The 2nd Workshop on Neural Architecture Search . Washington DC : ICLR , 2021 : 1 - 15 .

FIGUEROA BARRAZA J , LÓPEZ DROGUETT E , RAMOS MARTINS M . FS-SCF network: Neural network interpretability based on counterfactual generation and feature selection for fault diagnosis [J ] . Expert Systems with Applications , 2024 , 237 : 121670 .

RESHEF D N , RESHEF Y A , FINUCANE H K , et al . Detecting novel associations in large data sets [J ] . Science , 2011 , 334 ( 6062 ): 1518 - 1524 .

DEB K , PRATAP A , AGARWAL S , et al . A fast and elitist multiobjective genetic algorithm: NSGA-II [J ] . IEEE Transactions on Evolutionary Computation , 2002 , 6 ( 2 ): 182 - 197 .

DANDOLO D , MASIERO C , CARLETTI M , et al . AcME: Accelerated model-agnostic explanations: Fast whitening of the machine-learning black box [J ] . Expert Systems with Applications , 2023 , 214 : 119115 .

KEANE M T , SMYTH B . Good Counterfactuals and Where to Find Them: A Case-based Technique for Generating Counterfactuals for Explainable AI (XAI) [M ] // Case-Based Reasoning Research and Development . Cham : Springer International Publishing , 2020 : 163 - 178 .

YANG H R , CHEN H X , ZHANG S X , et al . Generating counterfactual hard negative samples for graph contrastive learning [C ] // Proceedings of the ACM Web Conference 2023 . New York : ACM , 2023 : 621 - 629 .

PEDREGOSA F , VAROQUAUX G , GRAMFORT A , et al . Scikit-learn: Machine learning in Python [J ] . Journal of Machine Learning Research , 2011 , 12 : 2825 - 2830 .

MARCEAU L , QIU L L , VANDEWIELE N , et al . A comparison of deep learning performances with other machine learning algorithms on credit scoring unbalanced data [EB/OL ] . ( 2020-02-25 )[ 2025-06-30 ] . https://arxiv.org/abs/1907.12363v2 https://arxiv.org/abs/1907.12363v2 .

BRUGHMANS D , LEYMAN P , MARTENS D . NICE: An algorithm for nearest instance counterfactual explanations [J ] . Data Mining and Knowledge Discovery , 2024 , 38 ( 5 ): 2665 - 2703 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Feature Selection of Heterogeneous Data Based on Multi-Objective Particle Swarm Optimization

Robust Unsupervised Feature Selection with Double Fuzzy Learning

Estimation of Signal-to-Noise Ratio Based on Feature Selection and Weight Optimization of High-Order Moments

Yaw Stability Control of Distributed Drive Electric Vehicle Based on Multi-Objective Parallel Chaos Optimization

Related Author

GONG Dun-wei

HU Ying

ZHANG Yong

GAO Yun-long

SHI Shu-guang

ZHAO Zhi-xiang

CAO Chao

PAN Jin-yan

Related Institution

School of Information and Electrical Engineering, China University of Mining and Technology

Tung Sah Institute of Micro-Nano Science and Technology, Xiamen University

National Institute for Data Science in Health and Medicine, Xiamen University

Third Institute of Oceanography, Ministry of Natural Resources

School of Ocean Information Engineering, Jimei University

⁰