Enhancing the Interpretability of Cardiovascular Disease Classifiers using Born-Again Tree Ensembles

  • L. G. S. N. A. Magalhães ENCE-IBGE
  • E. C. Gonçalves ENCE-IBGE

Resumo


According to the Pan American Health Organization, more people die each year from cardiovascular diseases than from any other cause. Due to this, ensemble classifiers such as Random Forest (RF) have been increasingly employed to build models targeted at the early prediction of such diseases. Nevertheless, one of the main disadvantages of the ensemble approaches lies in the fact that they cannot be applied when the goal is to build interpretable models (which are often desired or even required by both physicians and patients). To tackle this problem, in this work we evaluate the use of Born-Again Tree Ensembles (BA), a recently proposed technique that corresponds to the the first exact algorithm that transforms an RF into a single decision tree. Experiments carried out on a dataset containing data about 1,417 patients show that BA was able to produce a model that is directly interpretable, while at the same time keeping the same predictive power of an RF model.

Palavras-chave: born-again tree, random forest, interpretability, cardiovascular disease classification, data mining

Referências

Asif, D., Bibi, M., Arif, M. S., and Mukheimer, A. Enhancing heart disease prediction through ensemble learning techniques with hyperparameter optimization. Algorithms 16 (6), 2023.

Berka, P., Rauch, J., and Tomečková, M. Lessons Learned from the ECML/PKDD Discovery Challenge on the Atherosclerosis Risk Factors Data. Computing and Informatics 26 (3): 329–344, 2007.

Bhatt, C. M., Patel, P., Ghetia, T., and Mazzeo, P. L. Effective heart disease prediction using machine learning techniques. Algorithms 16 (88), 2023.

Breiman, L. Random Forests. Machine Learning vol. 45, pp. 5–32, 2001. Breiman, L., Friedman, J., Stone, C. J., and Olshen, R. A. Classification and Regression Trees. Taylor & Francis, 1984.

da Silva, V. R. and Gonçalves, E. C. Análise Comparativa de Implementações de Algoritmos de Árvores de Decisão para Aplicações no Serviço Público. In Anais da Escola Regional de Computação Bahia, Alagoas e Sergipe (ERBASE). Maceió, Brazil, pp. 10–19, 2021.

de Oliveira, G. M. M., Brant, L. C. C., Polanczyk, C. A., Malta, D. C., Biolo, A., Nascimento, B. R., de Fatima Marinho de Souza, M., Lorenzo, A. R. D., de Paiva Fagundes Júnior, A. A., Schaan, B. D., de Souza e Silva, C. G., de Castilho, F. M., Cesena, F. H. Y., Soares, G. P., Junior, G. F. X., Filho, J. A. S. B., Passaglia, L. G., Filho, M. M. P., Machline-Carrion, M. J., Bittencourt, M. S., Neto, O. M. P., Villela, P. B., Teixeira, R. A., Stein, R., Sampaio, R. O., Gaziano, T. A., Perel, P., Roth, G. A., and Ribeiro, A. L. P. Estatística Cardiovascular - Brasil. Arq. Bras. Cardiol. 121 (2): 1–131, 2024.

Doshi-Velez, F. and Kim, B. Towards A Rigorous Science of Interpretable Machine Learning. arXiv: Machine Learning, 2017.

ECML-PKDD. ECML/PKDD2004 Discovery Challenge. [link], 2004.

Freitas, A. A. Comprehensible Classification Models: A Position Paper. SIGKDD Explor. Newsl. 15 (1), mar, 2014. Gonçalves, E. C. and Plastino, A. Mining Strong Associations and Exceptions in the Stulong Data Set. In Proceedings of the ECML/PKDD Discovery Challenge. Pisa, Italy, pp. 44–55, 2004.

OPAS. Calculadora de risco cardiovascular. [link], 2024a.

OPAS. Organiza cão Pan-Americana da Saúde: Doen cas Cardiovasculares. [link], 2024b.

Rudin, C. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Nature Machine Intelligence vol. 1, pp. 206–215, 2019.

SCORE2 WORKING GROUP. SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. European Heart Journal 42 (25): 2439–2454, 2021.

Soulet, A. and Hébert, C. Using Emerging Patterns from Clusters to Characterize Social Subgroups of Patients Affected by Atherosclerosis. In Proceedings of the ECML/PKDD Discovery Challenge. Pisa, Italy, pp. 104–115, 2004.

Sumwiza, K., Twizere, C., Rushingabigwi, G., Bakunzibake, P., and Bamurigire, P. Enhanced Cardiovascular Disease Prediction Model Using Random Forest Algorithm. Informatics in Medicine Unlocked vol. 41, pp. 101316, 2023.

Vidal, T., Pacheco, T., and Schiffer, M. Born-Again Tree Ensembles. arXiv vol. arXiv:2003.11132, 2020.

Vidal, T. and Schiffer, M. Born-Again Tree Ensembles. In Proceedings of the 37th International Conference on Machine Learning, 2020.
Publicado
17/11/2024
MAGALHÃES, L. G. S. N. A.; GONÇALVES, E. C.. Enhancing the Interpretability of Cardiovascular Disease Classifiers using Born-Again Tree Ensembles. In: SYMPOSIUM ON KNOWLEDGE DISCOVERY, MINING AND LEARNING (KDMILE), 12. , 2024, Belém/PA. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2024 . p. 9-16. ISSN 2763-8944. DOI: https://doi.org/10.5753/kdmile.2024.243749.