Abstract
Modeling from data usually has two distinct facets: building sound explanatory models or creating powerful predictive models for a system or phenomenon. Most of recent literature does not exploit the relationship between explanation and prediction while learning models from data. Recent algorithms are not taking advantage of the fact that many phenomena are actually defined by diverse sub-populations and local structures, and thus there are many possible predictive models providing contrasting interpretations or competing explanations for the same phenomenon. In this article, we propose to explore a complementary link between explanation and prediction. Our main intuition is that models having their decisions explained by the same factors are likely to perform better predictions for data points within the same local structures. We evaluate our methodology to model the evolution of pain relief in patients suffering from chronic pain under usual guideline-based treatment. The ensembles generated using our framework are compared with all-in-one approaches of robust algorithms to high-dimensional data, such as Random Forests and XGBoost. Chronic pain can be primary or secondary to diseases. Its symptomatology can be classified as nociceptive, nociplastic, or neuropathic, and is generally associated with many different causal structures, challenging typical modeling methodologies. Our data includes 631 patients receiving pain treatment. We considered 338 features providing information about pain sensation, socioeconomic status, and prescribed treatments. Our goal is to predict, using data from the first consultation only, if the patient will be successful in treatment for chronic pain relief. As a result of this work, we were able to build ensembles that are able to consistently improve performance by up to 33% when compared to models trained using all the available features. We also obtained relevant gains in interpretability, with resulting ensembles using only 15% of the total number of features. We show we can effectively generate ensembles from competing explanations, promoting diversity in ensemble learning and leading to significant gains in accuracy by enforcing a stable scenario in which models that are dissimilar in terms of their predictions are also dissimilar in terms of their explanation factors.
- M. Abad-Grau, J. Ierache, C. Cervino, and P. Sebastiani. 2008. Evolution and challenges in the design of computational systems for triage assistance. J. Biomed. Inform. 41, 3 (2008), 432–441. Google Scholar
Digital Library
- R. Agrawal, J. Gehrke, D. Gunopulos, and P. Raghavan. 2005. Automatic subspace clustering of high dimensional data. Data Mining Knowl. Discov. 11, 1 (2005), 5–33. Google Scholar
Digital Library
- J. M. Benitez, J. L. Castro, and I. Requena. 1997. Are artificial neural networks black boxes?IEEE Trans. Neural Netw. 8, 5 (Sept. 1997), 1156–1164. DOI:https://doi.org/10.1109/72.623216 Google Scholar
Digital Library
- L. Bernardes, M. Carvalho, S. Harnik, M. Teixeira, J. Ottolia, D. Castro, A. Veloso, R. Francisco, C. Listik, R. Galhardoni, V. da Silva, L. Moreira, A. de Amorim Filho, A. Fernandes, and D. Ciampi de Andrade. 2021. Sorting pain out of salience: Assessment of pain facial expressions in the human fetus. Pain Rep. 6, 1 (2021), e882.Google Scholar
- F. Blyth, L. March, A. Brnabic, L. Jorm, M. Williamson, and M. Cousins. 2001. Chronic pain in Australia: A prevalence study. Pain 89, 2-3 (2001), 127–134. DOI:https://doi.org/10.1016/S0304-3959(00)00355-9Google Scholar
Cross Ref
- L. Breiman. 2001. Random forests. Mach. Learn. 45, 1 (2001), 5–32. Google Scholar
Digital Library
- M. Chen, K. Weinberger, and Y. Chen. 2011. Automatic feature decomposition for single view co-training. In International Conference on Machine Learning. 953–960. Google Scholar
Digital Library
- T. Chen and C. Guestrin. 2016. XGBoost: A scalable tree boosting system. In International Conference on Knowledge Discovery and Data Mining. 785–794. Google Scholar
Digital Library
- Y. Cheng and G. Church. 2000. Biclustering of expression data. In Proceedings of the 8th International Conference on Intelligent Systems for Molecular Biology. 93–103. Google Scholar
Digital Library
- R. Dworkin, D. Turk, K. Wyrwich, D. Beaton, D. Cleeland, J. Farrar, J. Haythornthwaite, M. Jensen, R. Kerns, D. Ader, N. Brandenburg, L. Burke, D. Cella, J. Chandler, P. Cowan P., R. Dimitrova, R. Dionne, S. Hertz, A. Jadad, N. Katz, H. Kehlet, L. Kramer, D. Manning, C. McCormick, M. McDermott, H. McQuay, S. Patel, L. Porter, S. Quessy, B. Rappaport, C. Rauschkolb, D. Revicki, and M. Rothman, 2008. Interpreting the clinical importance of treatment outcomes in chronic pain clinical trials: IMMPACT recommendations. J. Pain 9, 2 (2008), 105–121.Google Scholar
- Radwa Elshawi, Youssef Sherif, Mouaz Al-Mallah, and Sherif Sakr. 2019. Interpretability in healthcare a comparative study of local machine learning interpretability techniques. In IEEE 32nd International Symposium on Computer-Based Medical Systems (CBMS’19). IEEE, 275–280.Google Scholar
- M. Ester, H. Kriegel, J. Sander, and X. Xu. 1996. A density-based algorithm for discovering clusters in large spatial databases with noise. In International Conference on Knowledge Discovery and Data Mining. 226–231. Google Scholar
Digital Library
- T. Fawcett. 2006. An introduction to ROC analysis. Pattern Recog. Lett. 27, 8 (2006), 861–874. Google Scholar
Digital Library
- K. Ferreira, T. Bastos, D. Ciampi, A. Silva, J. Appolinario, M. Jacobsen, and M. Latorre. 2016. Prevalence of chronic pain in a metropolitan area of a developing country: A population-based study. Arquivos de Neuro-psiquiatria 74, 12 (2016), 990–998.Google Scholar
- A. Friesen and P. Domingos. 2015. Recursive decomposition for nonconvex optimization. In International Joint Conference on Artificial Intelligence. 253–259. Google Scholar
Digital Library
- B. Goldstein, A. Navar, and R. Carter. 2016. Moving beyond regression techniques in cardiovascular risk prediction: Applying machine learning to address analytic challenges. Eur. Heart J. 38, 23 (2016), 1805–1814.Google Scholar
- J. Hanley and B. McNeil. 1982. The meaning and use of the area under a receiver operating characteristic ROC curve. Radiology 143 (1982), 29–36.Google Scholar
Cross Ref
- J. Hill, K. Dunn, M. Lewis, R. Mullis, C. Main, N. Foster, and E. Hay. 2008. A primary care back pain screening tool: Identifying patient subgroups for initial treatment. Arthr. Rheum 5, 59 (2008), 632–641.Google Scholar
- Yan Huang, Huiru Zheng, Chris Nugent, Paul McCullagh, Norman Black, Kevin E. Vowles, and Lance McCracken. 2010. Feature selection and classification in supporting report-based self-management for people with chronic pain. IEEE Trans. Inf. Technol. Biomed. 15, 1 (2010), 54–61. IEEE. Google Scholar
Digital Library
- D. Jha and G. Kwon. 2017. Diagnosis of Alzheimer’s disease using a machine learning technique. Alzh. Dement. 13, 7 (2017), 1538.Google Scholar
- L. Kuncheva and C. Whitaker. 2003. Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy. Mach. Learn. 51, 2 (2003), 181–207. Google Scholar
Digital Library
- Ludmila I. Kuncheva, Fabio Roli, Gian Luca Marcialis, and Catherine A. Shipp. 2001. Complexity of data subsets generated by the random subspace method: An experimental investigation. In International Workshop on Multiple Classifier Systems. Springer, 349–358. Google Scholar
Digital Library
- S. Lundberg and S. Lee. 2017. A unified approach to interpreting model predictions. In Conference on Neural Information Processing Systems. 4768–4777. Google Scholar
Digital Library
- S. Lundberg, B. Nair, M. Vavilala, M. Horibe, M. Eisses, T. Adams, D. Liston, D. Low, S. Newman, J. Kim, and S. Lee. 2018. Explainable machine learning predictions to help anesthesiologists prevent hypoxemia during surgery. Nat. Biomed. Eng. 2, 10 (2018), 749–760. DOI:https://doi.org/10.1101/206540Google Scholar
Cross Ref
- Scott M. Lundberg, Gabriel Erion, Hugh Chen, Alex DeGrave, Jordan M. Prutkin, Bala Nair, Ronit Katz, Jonathan Himmelfarb, Nisha Bansal, and Su-In Lee. 2020. From local explanations to global understanding with explainable AI for trees. Nat. Mach. Intell. 2, 1 (Jan. 2020), 56–67. DOI:https://doi.org/10.1038/s42256-019-0138-9Google Scholar
Cross Ref
- O. Maimon and L. Rokach. 2002. Improving supervised learning by feature decomposition. In International Symposium on the Foundations of Information and Knowledge Systems. 178–196. Google Scholar
Digital Library
- Morteza Mashayekhi and Robin Gras. 2015. Rule extraction from random forest: The RF+HC methods. In Advances in Artificial Intelligence, Denilson Barbosa and Evangelos Milios (Eds.). Vol. 9091. Springer International Publishing, Cham, 223–237. DOI:https://doi.org/10.1007/978-3-319-18356-5_20Google Scholar
- G. McLachlan and D. Peel. 2000. Finite Mixture Models. Wiley Series in Probability and Statistics. John Wiley & Sons, New York.Google Scholar
- R. Melzack. 1975. The McGill pain questionnaire. major properties and scoring methods. Pain 1 (1975), 277–299.Google Scholar
Cross Ref
- A. Navani and G. Li. 2016. Chronic pain challenge: A statistical machine-learning method for chronic pain assessment. J. Rec. Adv. Pain 2, 3 (2016), 82–86. DOI:https://doi.org/10.5005/jp-journals-10046-0048Google Scholar
- E. Nigri, N. Ziviani, F. Cappabianco, A. Antunes, and A. Veloso. 2020. Explainable deep CNNs for MRI-based diagnosis of Alzheimer’s disease. In International Joint Conference on Neural Networks. IEEE, 1–8.Google Scholar
- A. Oliveira and S. Madeira. 2004. Biclustering algorithms for biological data analysis: A survey. IEEE/ACM Trans. Computat. Biol. Bioinf. 1, 1 (2004), 24–45. Google Scholar
Digital Library
- T. Pansombut, W. Hendrix, Z. Jacob Gao, B. Harrison, and N. Samatova. 2011. Biclustering-driven ensemble of Bayesian belief network classifiers for underdetermined problems. In International Joint Conference on Artificial Intelligence. 1439–1445. Google Scholar
Digital Library
- F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. VanderPlas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. 2011. Scikit-learn: Machine learning in Python. J. Mach. Learn. Res. 12 (2011), 2825–2830. Google Scholar
Digital Library
- A. Pieterse, A. Stiggelbout, and V. Montori. 2019. Shared decision making and the importance of time. JAMA - J. Amer. Med. Assoc. 322, 1 (2019), 25–26. DOI:https://doi.org/10.1001/jama.2019.3785Google Scholar
- N. Pombo, P. Araújo, and J. Viana. 2014. Knowledge discovery in clinical decision support systems for pain management: A systematic review. Arti. Intell. Med. 60, 1 (2014), 1–11. Google Scholar
Digital Library
- M. Ribeiro, S. Singh, and C. Guestrin. 2016. “Why should i trust you?”: Explaining the predictions of any classifier. In International Conference on Knowledge Discovery and Data Mining. 1135–1144. Google Scholar
Digital Library
- M. Ribeiro, S. Singh, and C. Guestrin. 2018. Anchors: High-precision model-agnostic explanations. In AAAI Conference on Artificial Intelligence. 1527–1535.Google Scholar
- Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. 2018. Anchors: High-precision model-agnostic explanations. In AAAI Conference on Artificial Intelligence, Vol. 18. 1527–1535.Google Scholar
- Greg Ridgeway, David Madigan, Thomas Richardson, and John O’Kane. 1998. Interpretable boosted naïve Bayes classification. In KDD. 101–104. Google Scholar
Digital Library
- Michael E. Robinson, Andrew M. O’Shea, Jason G. Craggs, Donald D. Price, Janelle E. Letzen, and Roland Staud. 2015. Comparison of machine classification algorithms for fibromyalgia: Neuroimages versus self-report. J. Pain 16, 5 (2015), 472–477. Elsevier.Google Scholar
Cross Ref
- G. Shmueli. 2010. To explain or to predict?Statist. Sci. 25, 3 (2010), 289–310.Google Scholar
- Martin Tamajka, Wanda Benesova, and Matej Kompanek. 2019. Transforming convolutional neural network to an interpretable classifier. In International Conference on Systems, Signals and Image Processing (IWSSIP’19). IEEE, 255–259. ZSCC: 0000002.Google Scholar
- A. Tanay, R. Sharan, M. Kupiec, and R. Shamir. 2004. Revealing modularity and organization in the yeast molecular network by integrated analysis of highly heterogeneous genomewide data. Proc. Natl. Acad. Sci. 101, 9 (2004), 2981–2986.Google Scholar
Cross Ref
- A. Tanay, R. Sharan, and R. Shamir. 2005. Biclustering algorithms: A survey. Handb. Computat. Molec. Biol. 9 (2005), 26–1.Google Scholar
- D. Valle, T. Pimentel, and A. Veloso. 2020. Assessing the reliability of visual explanations of deep models with adversarial perturbations. In International Joint Conference on Neural Networks. IEEE, 1–8.Google Scholar
- L. van der Maaten. 2009. Learning a parametric embedding by preserving local structure. In International Conference on Artificial Intelligence and Statistics. 384–391.Google Scholar
- T. Vos, A. Flaxman, M. Naghavi, R. Lozano, C. Michaud, M. Ezzati, K. Shibuya, J. Salomon, S. Abdalla, and V. Aboyans. 2012. Years lived with disability (YLDs) for 1160 sequelae of 289 diseases and injuries 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet 380, 9859 (2012), 2163–2196.Google Scholar
- J. Ward. 1963. Hierarchical grouping to optimize an objective function. J. Amer. Statist. Assoc. 58 (1963), 236–244.Google Scholar
Cross Ref
- K. White, J. Lee, and A. de C Williams. 2016. Are patients’ and doctors’ accounts of the first specialist consultation for chronic back pain in agreement?J. Pain Res. (2016), 1109–1120.Google Scholar
- A. Williams and K. Craig. 2016. Updating the definition of pain. PAIN 157 (05 2016), 1. DOI:https://doi.org/10.1097/j.pain.0000000000000613Google Scholar
Index Terms
Predicting the Evolution of Pain Relief: Ensemble Learning by Diversifying Model Explanations
Recommendations
Predicting Major Adverse Kidney Events among Critically Ill Adults Using the Electronic Health Record
Prediction of major adverse kidney events in critically ill patients may help target therapy, allow risk adjustment, and facilitate the conduct of clinical trials. In a cohort comprised of all critically ill adults admitted to five intensive care units ...
Predicting overall survivability in comorbidity of cancers
Cancer and other chronic diseases have constituted (and will do so at an increasing pace) a significant portion of healthcare costs in the United States in recent years. Although prior research has shown that diagnostic and treatment recommendations ...
Predicting readmission risk with institution-specific prediction models
ObjectiveThe ability to predict patient readmission risk is extremely valuable for hospitals, especially under the Hospital Readmission Reduction Program of the Center for Medicare and Medicaid Services which went into effect starting October 1, 2012. ...






Comments