An Optimized Ensemble Machine Learning Framework with Automated Feature Engineering for Multiclass Survival Prediction in Primary Biliary Cirrhosis
Main Article Content
Abstract
Primary Biliary Cholangitis is an autoimmune cholestatic disease that exhibits a progressively heterogeneous nature of its outcome. While establishing a static three-class survival target for predicting the progression of the disease offers a significant baseline framework for computation, the prediction is seriously hampered by the extremely skewed class distribution where transplantations are only 6.0%, while lab records are 32.5% missing. To address these constraints within an unconfounded historical natural history baseline, an integrated machine learning pipeline was evaluated on the Mayo Clinic dataset spanning a primary cohort of 418 patients and a validation cohort of 312 patients. Missing data fields were recovered via a tree-based Multiple Imputation by Chained Equations framework driven by CatBoost base learners to preserve multi-dimensional covariance structures, followed by an automated feature engineering module that extracted orthogonal arithmetic interactions. A heterogeneous ensemble of 21 base classifiers was optimized through a continuous Optuna CMA-ES Bayesian search strategy evaluated under a 5-fold stratified cross-validation paradigm. The optimized framework achieved a mean log loss of 0.469 ± 0.157 and a micro-averaged F1-score of 0.842 ± 0.048, matching standalone ExtraTrees accuracy (0.854 ± 0.024; t = -0.679, p = 0.535) but functioning as the unique cost-sensitive architecture capable of overcoming majoritarian class collapse to achieve non-zero transplantation sensitivity at a localized recall of 0.200. This non-zero local sensitivity operates as a vital early-warning clinical triage alert, demonstrating that continuous ensemble fusion successfully mitigates highly skewed multi-outcome boundaries before transitioning toward dynamic time-to-event competing-risk workflows.
Article Details
This work is licensed under a Creative Commons Attribution 4.0 International License.
References
[1] K. D. Lindor, C. L. Bowlus, J. Boyer, C. Levy, and M. Mayo, “Primary biliary cholangitis: 2021 practice guidance update from the American Association for the Study of Liver Diseases,” Hepatology, vol. 75, no. 4, pp. 1012–1013, Apr. 2022, doi: 10.1002/hep.32117.
[2] A. Gerussi et al., “Machine learning in primary biliary cholangitis: A novel approach for risk stratification,” Liver International, vol. 42, no. 3, pp. 615–627, Mar. 2022, doi: 10.1111/liv.15141.
[3] M. Carbone et al., “The UK‐PBC risk scores: Derivation and validation of a scoring system for long‐term prediction of end‐stage liver disease in primary biliary cholangitis,” Hepatology, vol. 63, no. 3, pp. 930–950, Mar. 2016, doi: 10.1002/hep.28017.
[4] the G. P. B. C. S. Group et al., “Prognostic scores for ursodeoxycholic acid-treated patients predict graft loss and mortality in recurrent primary biliary cholangitis after liver transplantation,” Journal of Hepatology, vol. 81, no. 4, pp. 679–689, Oct. 2024, doi: 10.1016/j.jhep.2024.05.010.
[5] J. P. Fine and R. J. Gray, “A Proportional Hazards Model for the Subdistribution of a Competing Risk,” Journal of the American Statistical Association, vol. 94, no. 446, pp. 496–509, Jun. 1999, doi: 10.1080/01621459.1999.10474144.
[6] J. Li et al., “Method for Incomplete and Imbalanced Data Based on Multivariate Imputation by Chained Equations and Ensemble Learning,” IEEE Journal of Biomedical and Health Informatics, vol. 28, no. 5, pp. 3102–3113, May 2024, doi: 10.1109/JBHI.2024.3376428.
[7] S. E. Müller et al., “Machine Learning Models predicting Decompensation in Cirrhosis,” Journal of Gastrointestinal and Liver Diseases, vol. 34, no. 1, pp. 71–80, Mar. 2025, doi: 10.15403/jgld-5876.
[8] W. R. Kim et al., “MELD 3.0: The Model for End-Stage Liver Disease Updated for the Modern Era,” Gastroenterology, vol. 161, no. 6, pp. 1887-1895.e4, Dec. 2021, doi: 10.1053/j.gastro.2021.08.050.
[9] P. J. Johnson et al., “Assessment of Liver Function in Patients With Hepatocellular Carcinoma: A New Evidence-Based Approach—The ALBI Grade,” Journal of Clinical Oncology, vol. 33, no. 6, pp. 550–558, Feb. 2015, doi: 10.1200/JCO.2014.57.9151.
[10] N. Kimura et al., “Development and validation of machine learning model for predicting treatment responders in patients with primary biliary cholangitis,” Hepatology Research, vol. 54, no. 1, pp. 67–77, Jan. 2024, doi: 10.1111/hepr.13966.
[11] N. Kimura et al., “Machine learning prediction model for treatment responders in patients with primary biliary cholangitis,” JGH Open, vol. 7, no. 6, pp. 431–438, Jun. 2023, doi: 10.1002/jgh3.12915.
[12] B. Da et al., “Development and validation of a machine learning-based model to predict survival in patients with cirrhosis after transjugular intrahepatic portosystemic shunt,” eClinicalMedicine, vol. 79, p. 103001, Jan. 2025, doi: 10.1016/j.eclinm.2024.103001.
[13] R. E. Dickson, P. M. Grambsch, T. R. Fleming, L. D. Fisher, and A. Langworthy, “Prognosis in primary biliary cirrhosis: Model for decision making,” Hepatology, vol. 10, no. 1, pp. 1–7, Jul. 1989, doi: 10.1002/hep.1840100102.
[14] I. R. White, P. Royston, and A. M. Wood, “Multiple imputation using chained equations: Issues and guidance for practice,” Statistics in Medicine, vol. 30, no. 4, pp. 377–399, Feb. 2011, doi: 10.1002/sim.4067.
[15] Haibo He and E. A. Garcia, “Learning from Imbalanced Data,” IEEE Transactions on Knowledge and Data Engineering, vol. 21, no. 9, pp. 1263–1284, Sep. 2009, doi: 10.1109/TKDE.2008.239.
[16] E. Hallaji, R. Razavi-Far, V. Palade, and M. Saif, “Adversarial Learning on Incomplete and Imbalanced Medical Data for Robust Survival Prediction of Liver Transplant Patients,” IEEE Access, vol. 9, pp. 73641–73650, 2021, doi: 10.1109/ACCESS.2021.3081040.
[17] A. Gerussi et al., “Trends in Primary Biliary Cholangitis: Prospective Cohort Study From the European Reference Network Registry (R‐LIVER),” United European Gastroenterology Journal, vol. 13, no. 10, pp. 1955–1963, Dec. 2025, doi: 10.1002/ueg2.70134.
[18] Y. Bengio, A. C. Courville, and P. Vincent, “Representation Learning: A Review and New Perspectives,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, pp. 1798–1828, 2012, [Online]. Available: https://api.semanticscholar.org/CorpusID:393948
[19] J. M. Shapiro, H. Smith, and F. Schaffner, “Serum bilirubin: a prognostic factor in primary biliary cirrhosis.,” Gut, vol. 20, no. 2, pp. 137–140, Feb. 1979, doi: 10.1136/gut.20.2.137.
[20] L. Ostroumova, G. Gusev, A. Vorobev, A. V. Dorogush, and A. Gulin, “CatBoost: unbiased boosting with categorical features,” in Neural Information Processing Systems, 2017. [Online]. Available: https://api.semanticscholar.org/CorpusID:5044218
[21] D. Sharma et al., “GraftIQ: Hybrid multi-class neural network integrating clinical insight for multi-outcome prediction in liver transplant recipients,” Nature Communications, vol. 16, no. 1, p. 4943, May 2025, doi: 10.1038/s41467-025-59610-8.
[22] F. Mohtasham, S. S. Hashemi Nazari, M. A. Pourhoseingholi, K. Kavousi, and M. R. Zali, “Hybrid feature-selection and diversity-guided stacking framework for interpretable ensemble learning: Application to COVID-19 mortality prediction,” PLOS One, vol. 21, no. 4, p. e0341198, Apr. 2026, doi: 10.1371/journal.pone.0341198.
[23] Z.-H. Zhou, Ensemble Methods. Boca Raton: Chapman and Hall/CRC, 2025. doi: 10.1201/9781003587774.
[24] T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama, “Optuna: A Next-generation Hyperparameter Optimization Framework,” in Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, New York, NY, USA: ACM, Jul. 2019, pp. 2623–2631. doi: 10.1145/3292500.3330701.
[25] J. M. Johnson and T. M. Khoshgoftaar, “Cost-Sensitive Ensemble Learning for Highly Imbalanced Classification,” in 2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA), IEEE, Dec. 2022, pp. 1427–1434. doi: 10.1109/ICMLA55696.2022.00225.
[26] M. Sokolova and G. Lapalme, “A systematic analysis of performance measures for classification tasks,” Information Processing & Management, vol. 45, no. 4, pp. 427–437, Jul. 2009, doi: 10.1016/j.ipm.2009.03.002.
[27] J. Waring, C. Lindvall, and R. Umeton, “Automated machine learning: Review of the state-of-the-art and opportunities for healthcare,” Artificial Intelligence in Medicine, vol. 104, p. 101822, Apr. 2020, doi: 10.1016/j.artmed.2020.101822.
[28] J. X. C. Ke, A. DhakshinaMurthy, R. B. George, and P. Branco, “The effect of resampling techniques on the performances of machine learning clinical risk prediction models in the setting of severe class imbalance: development and internal validation in a retrospective cohort,” Discover Artificial Intelligence, vol. 4, no. 1, p. 91, Nov. 2024, doi: 10.1007/s44163-024-00199-0.