當前研究顯示異質性治療效應和條件平均治療效應在精準醫療和醫療資源配置中的關鍵重要性，促進個體治療效果的多樣估計。然而，普遍的次群體分析面臨挑戰：採用黑盒模型具有良好預測準確性但尚失解釋性，或提供可解釋的結果但預測能力有限。為解決這種差異，我們提出了一種新方法，稱為因果規則森林，結合了預測準確性和可解釋性。通過創新的建模技術和算法，我們的方法提高了次群體分析的效率，同時保持了解釋性並實現了超越複雜模型的性能，從而擴大了其應用範圍。通過實驗，我們展示了我們的方法在複雜的因果推論任務中的靈活性和效能，提供了有價值的治療效果，並以易於閱讀和理解的格式提取規則，增強模型的說服力。

Current research demonstrates the critical importance of Heterogeneous Treatment Effects (HTE) and Conditional Average Treatment Effects (CATE) in personalized medicine and healthcare resource allocation, facilitating diverse estimations of individual treatment effects (ITE). However, prevalent subgroup analyses face challenges: employing black-box models yields strong predictive accuracy but lacks interpretability, or provides interpretable outcomes with limited predictive capability. To address this disparity, we propose a novel method, termed Causal Rule Forest, which integrates predictive accuracy and interpretability. Through innovative modeling techniques and algorithms, our method enhances the efficiency of subgroup analysis while maintaining interpretability and surpassing the performance of complex models, thereby expanding its scope of application. Through experimentation, we demonstrate the flexibility and efficacy of our method in complex causal inference tasks, providing valuable insights into treatment effects, and extracting rules in a format that is easy to read and comprehend, thereby enhancing the persuasiveness of the model.

論文審定書 i
摘要 ii
Abstract iii
List of Figures v
List of Table vi
1 Introduction 1
2 Background and Related Work 4
2.1 The Framework of Causal Inference 4
2.1.1 Potential Outcomes Framework (Rubin) 5
2.1.2 Structural Causal Models (SCM) Framework (Pearl) 7
2.2 Estimating Subgroup Analysis 9
2.3 Treatment Effect 14
2.4 Bayesian Additive Regression Trees (BART) 19
3 Estimating Treatment Effect Using Causal Rule Forest 22
3.1 Problem Setup 22
3.2 Building Causal Rule Forest 24
4 Experiment 29
4.1 Experiment Setup 29
4.2 Performance on Treatment Effect Estimation 31
4.3 Rule Interpretation with Causal Rule Forest 36
5 Conclusion and Discussion 40
Reference 42

Alaa, A. M. (n.d.). Limits of Estimating Heterogeneous Treatment Effects: Guidelines for Practical Algorithm Design.
Athey, S., & Imbens, G. (2016). Recursive Partitioning for Heterogeneous Causal Effects. Proceedings of the National Academy of Sciences, 113(27), 7353–7360. https://doi.org/10.1073/pnas.1510489113
Athey, S., Tibshirani, J., & Wager, S. (2019). Generalized random forests. The Annals of Statistics, 47(2). https://doi.org/10.1214/18-AOS1709
Bica, I., Alaa, A. M., Lambert, C., & Van Der Schaar, M. (2021). From Real‐World Patient Data to Individualized Treatment Effects Using Machine Learning: Current and Future Methods to Address Underlying Challenges. Clinical Pharmacology & Therapeutics, 109(1), 87–100. https://doi.org/10.1002/cpt.1907
Chipman, H. A., George, E. I., & McCulloch, R. E. (2010). BART: Bayesian additive regression trees. The Annals of Applied Statistics, 4(1). https://doi.org/10.1214/09-AOAS285
Curth, A., Svensson, D., & Weatherall, J. (n.d.). Really Doing Great at Estimating CATE? A Critical Look at ML Benchmarking Practices in Treatment Effect Estimation.
Gross, R. T. (n.d.). Infant Health and Development Program (IHDP): Enhancing the Outcomes of Low Birth Weight, Premature Infants in the United States, 1985-1988 (ICPSR 9795). Inter-university Consortium for Political and Social Research. https://doi.org/10.3886/ICPSR09795.v2
Hill, J. L. (2011). Bayesian Nonparametric Modeling for Causal Inference. Journal of Computational and Graphical Statistics, 20(1), 217–240. https://doi.org/10.1198/jcgs.2010.08162
Hiraishi, M., Wan, K., Tanioka, K., Yadohisa, H., & Shimokawa, T. (2023). Causal rule ensemble method for estimating heterogeneous treatment effect with consideration of main effects (arXiv:2307.14766). arXiv. http://arxiv.org/abs/2307.14766
Johansson, F. D., Shalit, U., Kallus, N., & Sontag, D. (2023). Generalization Bounds and Representation Learning for Estimation of Potential Outcomes and Causal Effects (arXiv:2001.07426). arXiv. http://arxiv.org/abs/2001.07426
Kang, Y., Huang, S.-T., & Wu, P.-H. (2021). Detection of Drug–Drug and Drug–Disease Interactions Inducing Acute Kidney Injury Using Deep Rule Forests. SN Computer Science, 2(4), 299. https://doi.org/10.1007/s42979-021-00670-0
Künzel, S. R., Sekhon, J. S., Bickel, P. J., & Yu, B. (2019). Meta-learners for Estimating Heterogeneous Treatment Effects using Machine Learning. Proceedings of the National Academy of Sciences, 116(10), 4156–4165. https://doi.org/10.1073/pnas.1804597116
Kuo, B., Kang, Y., Wu, P., Huang, S.-T., & Huang, Y. (2020). Discovering Drug-Drug and Drug-Disease Interactions Inducing Acute Kidney Injury Using Deep Rule Forests. 2020 IEEE 21st International Conference on Information Reuse and Integration for Data Science (IRI), 385–390. https://doi.org/10.1109/IRI49571.2020.00062
Lalonde, R. J. (2024). Evaluating the Econometric Evaluations of Training Programs with Experimental Data.
Matthews, J. N. S. (2006). Introduction to Randomized Controlled Clinical Trials (0 ed.). Chapman and Hall/CRC. https://doi.org/10.1201/9781420011302
NPCI package. (n.d.). https://github.com/vdorie/npci
Pearl, J. (n.d.). Causality: Models, Reasoning and Inference (2nd Edition). Cambridge University Press.
Pearl, J. (2009). Causal inference in statistics: An overview. Statistics Surveys, 3(none). https://doi.org/10.1214/09-SS057
Rubin, D. B. (1990). [On the Application of Probability Theory to Agricultural Experiments. Essay on Principles. Section 9.] Comment: Neyman (1923) and Causal Inference in Experiments and Observational Studies. Statistical Science, 5(4), 472–480. https://doi.org/10.1214/ss/1177012032
Rubin, D. B. (2005). Causal Inference Using Potential Outcomes: Design, Modeling, Decisions. Journal of the American Statistical Association, 100(469), 322–331. https://doi.org/10.1198/016214504000001880
Rudin, C., Chen, C., Chen, Z., Huang, H., Semenova, L., & Zhong, C. (2021). Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (arXiv:2103.11251). arXiv. http://arxiv.org/abs/2103.11251
Shalit, U., Johansson, F. D., & Sontag, D. (n.d.). Estimating individual treatment effect: Generalization bounds and algorithms.
Su, X., Kang, J., Fan, J., Levine, R. A., & Yan, X. (n.d.). Facilitating Score and Causal Inference Trees for Large Observational Studies.
Su, X., Tsai, C.-L., Wang, H., Nickerson, D. M., & Li, B. (2009). Subgroup Analysis via Recursive Partitioning. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.1341380
Wu, A., Yuan, J., Kuang, K., Li, B., Wu, R., Zhu, Q., Zhuang, Y. T., & Wu, F. (2022). Learning Decomposed Representations for Treatment Effect Estimation. IEEE Transactions on Knowledge and Data Engineering, 1–1. https://doi.org/10.1109/TKDE.2022.3150807
Wu, Y., Liu, H., Ren, K., & Chang, X. (2023). Causal Rule Learning: Enhancing the Understanding of Heterogeneous Treatment Effect via Weighted Causal Rules (arXiv:2310.06746). arXiv. http://arxiv.org/abs/2310.06746
Yoon, J., & Jordon, J. (2018). GANITE: ESTIMATION OF INDIVIDUALIZED TREAT- MENT EFFECTS USING GENERATIVE ADVERSARIAL.