急性腎損傷是正常的腎臟受到某種原因的傷害，導致腎功能突然減少，通常在數小時至幾週內發生，這會使得腎臟清除體內毒素及水份的能力在短時間內急劇下降。可能的病因包括慢性腎衰竭，藥物之間的交互作用，腎毒性藥物過量，例如：非類固醇類消炎藥(NSAIDs)。得到急性腎損傷的患者致死率高達60%，因此，如果能早期確診，早期治療，避免其衍生的併發症，可增加腎功能恢復的機會，最終提高患者的生存率，以及降低醫療成本。而藥物的交互作用是導致急性腎損傷的原因之一，藥物的交互作用是指當藥物與第二種藥物一起使用時，藥物對身體的影響發生變化，藥物相互作用可以延遲，減少或增強任一種藥物的吸收或引起藥物不良反應。我們透過台灣健保資料庫的曾經得過急性腎損傷的病人資料去發掘更多疾病藥物組合，因此，研究目的為找出什麼樣的疾病和藥物的排列組合可能導致急性腎損傷，我們提出了一種機器學習算法，即深度規則森林（Deep Rule Forest, DRF），它有助於從樹的模型中發現和提取規則，因為藥物和疾病的使用組合可以幫助識別上述相互作用透過深度規則森林所找出的疾病藥物規則，符合文獻也又找出未曾發掘過的規則，而在準確度方面DRF 模型在正確率的表現比傳統單一樹的方法與線性模式好。我們也證明層數越深，得到的規則也越精準。

Patients with kidney diseases are often diagnosed with Acute Kidney Injury (AKI). The mortality rate of critically ill patients with AKI is 60%. As a result, if AKI is diagnosed earlier, patients may have greater chances to recover renal function, which will ultimately improve the patients’ survival rate. The risk factors to AKI include drug-drug interactions and drug-disease interactions. According to previous researches, researchers used statistical analysis to measure the correlations between one disease and one drug. However, realistically, the correlations can be various when the patients usually have many prescriptions and complications. In this thesis, we propose a machine learning algorithm, Deep Rule Forests (DRF), which helps discover and extract rules from tree models as the combinations of drug and diseases usages to help identify aforementioned interactions. We also found that several drug and diseases usages that may be considered having significant impact on (re)occurrence of AKI. After that, the results show that DRF model performs better than typical tree-based and linear method in terms of the prediction accuracy. Moreover, we can obtain a series of situations that may cause AKI. If the layer of DRF model is higher, the extracted rules are more precise.

論文審訂書 i
Acknowledgement ii
中文摘要 iii
Abstract iv
1 Introduction 1
2 Literature Review 4
2.1 Acute Kidney Injury 4
2.2 Drug-induced renal toxicity 6
2.3 Drug-drug Interactions 7
2.4 Adverse Drug Reaction 8
2.5 Tree-based method 9
2.6 Representation learning and deep architectures 11
3 Detection of DDIs for AKI—Using Deep Rule Forests 12
3.1 Deep Rule Forests 12
3.2 Elastic net 14
4 Experimental Result 16
4.1 Data source 16
4.2 Compare the tree-based method, Lasso, and DRF model 16
4.3 Compare the rules between CART and DRF model 18
4.4 Display high risk group with AKI in forest plot 20
5 Discussion 24
6 Conclusion 25
7 Reference 26

Administration, N. H. I. (2014). Ministry of Health and Welfare. Ministry of Health and Welfare, Taiwan, R.O.C., 2014-2015.
Bagshaw, S. M., George, C., & Bellomo, R. (2008). Early acute kidney injury and sepsis: a multicentre evaluation. Critical care, 12(2), R47.
Bellomo, R., Kellum, J. A., & Ronco, C. (2012). Acute kidney injury. The Lancet, 380(9843), 756-766.
Bengio, Y. (2009). Learning deep architectures for AI. Foundations and trends® in Machine Learning, 2(1), 1-127.
Bengio, Y., Courville, A., & Vincent, P. (2013). Representation learning: A review and new perspectives. IEEE transactions on pattern analysis and machine intelligence, 35(8), 1798-1828.
Calderón-Larrañaga, A., Poblador-Plou, B., González-Rubio, F., Gimeno-Feliu, L. A., Abad-Díez, J. M., & Prados-Torres, A. (2012). Multimorbidity, polypharmacy, referrals, and adverse drug events: are we doing things well? Br J Gen Pract, 62(605), e821-e826.
Chawla, L. S., Eggers, P. W., Star, R. A., & Kimmel, P. L. (2014). Acute kidney injury and chronic kidney disease as interconnected syndromes. New England Journal of Medicine, 371(1), 58-66.
De Bartolo, L. (2016). Acute Kidney Injury (AKI). Encyclopedia of Membranes, 7-7.
Dormuth, C. R., Hemmelgarn, B. R., Paterson, J. M., James, M. T., Teare, G. F., Raymond, C. B., . . . Ernst, P. (2013). Use of high potency statins and rates of admission for acute kidney injury: multicenter, retrospective observational analysis of administrative databases. Bmj, 346, f880.
Finlay, S., & Jones, M. C. (2017). Acute kidney injury. Medicine.
Girardeau, Y., Trivin, C., Durieux, P., Le Beller, C., Neuraz, A., Degoulet, P., & Avillach, P. (2015). Detection of drug–drug interactions inducing acute kidney injury by electronic health records mining. Drug safety, 38(9), 799-809.
Hsu, C.-Y., McCulloch, C., Fan, D., Ordonez, J., Chertow, G., & Go, A. (2007). Community-based incidence of acute renal failure. Kidney international, 72(2), 208-212.
Izzedine, H., Launay-Vacher, V., & Deray, G. (2005). Antiviral drug-induced nephrotoxicity. American Journal of Kidney Diseases, 45(5), 804-817.
James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An introduction to statistical learning (Vol. 112): Springer.
Juurlink, D. N., Mamdani, M., Kopp, A., Laupacis, A., & Redelmeier, D. A. (2003). Drug-drug interactions among elderly patients hospitalized for drug toxicity. Jama, 289(13), 1652-1658.
Kane-Gill, S. L., Visweswaran, S., Saul, M. I., Wong, A.-K. I., Penrod, L. E., & Handler, S. M. (2011). Computerized detection of adverse drug reactions in the medical intensive care unit. International journal of medical informatics, 80(8), 570-578.
Kuo, C.-C., Chou, Y.-H., Lee, P.-H., & Chen, C.-H. (2009). Recent Update on Acute Kidney Injury and Critical Dialysis. Journal of Internal Medicine of Taiwan, 20(4), 320-334.
Lachance, P., Villeneuve, P.-M., Rewa, O. G., Wilson, F. P., Selby, N. M., Featherstone, R. M., & Bagshaw, S. M. (2017). Association between e-alert implementation for detection of acute kidney injury and outcomes: a systematic review. Nephrology Dialysis Transplantation, 32(2), 265-272.
Loboz, K. K., & Shenfield, G. M. (2005). Drug combinations and impaired renal function–the ‘triple whammy’. British journal of clinical pharmacology, 59(2), 239-243.
Luyckx, V. A., & Naicker, S. (2008). Acute kidney injury associated with the use of traditional medicines. Nature Clinical Practice Nephrology, 4(12), 664-671.
McCoy, A. B., Waitman, L. R., Gadd, C. S., Danciu, I., Smith, J. P., Lewis, J. B., . . . Peterson, J. F. (2010). A computerized provider order entry intervention for medication safety during acute kidney injury: a quality improvement report. American Journal of Kidney Diseases, 56(5), 832-841.
Mehta, R. L., Kellum, J. A., Shah, S. V., Molitoris, B. A., Ronco, C., Warnock, D. G., & Levin, A. (2007). Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Critical care, 11(2), R31.
Mehta, R. L., Pascual, M. T., Soroko, S., Savage, B. R., Himmelfarb, J., Ikizler, T. A., . . . Disease, P. t. I. C. i. A. R. (2004). Spectrum of acute renal failure in the intensive care unit: the PICARD experience. Kidney international, 66(4), 1613-1621.
Miller, K., Hettinger, C., Humpherys, J., Jarvis, T., & Kartchner, D. (2017). Forward Thinking: Building Deep Random Forests. arXiv preprint arXiv:1705.07366.
Ming-Hong Chen, Y.-C. K., and Pei-Wen Wang. (2009). Possible Mechanism of Contraindication to Acarbose in Type 2 Diabetic Patients with Renal Failure. J Intern Med Taiwan(20), 434-439.
Pannu, N., & Nadim, M. K. (2008). An overview of drug-induced acute kidney injury. Critical care medicine, 36(4), S216-S223.
Perazella, M. A. (2003). Drug-induced renal failure: update on new medications and unique mechanisms of nephrotoxicity. The American journal of the medical sciences, 325(6), 349-362.
Pirmohamed, M., James, S., Meakin, S., Green, C., Scott, A. K., Walley, T. J., . . . Breckenridge, A. M. (2004). Adverse drug reactions as cause of admission to hospital: prospective analysis of 18 820 patients. Bmj, 329(7456), 15-19.
Qato, D. M., Wilder, J., Schumm, L. P., Gillet, V., & Alexander, G. C. (2016). Changes in prescription and over-the-counter medication and dietary supplement use among older adults in the United States, 2005 vs 2011. JAMA internal medicine, 176(4), 473-482.
Tannenbaum, C., & Sheehan, N. L. (2014). Understanding and preventing drug–drug and drug–gene interactions. Expert review of clinical pharmacology, 7(4), 533-544.
Uchino, S., Kellum, J. A., Bellomo, R., Doig, G. S., Morimatsu, H., Morgera, S., . . . Macedo, E. (2005). Acute renal failure in critically ill patients: a multinational, multicenter study. Jama, 294(7), 813-818.
Ungprasert, P., Cheungpasitporn, W., Crowson, C. S., & Matteson, E. L. (2015). Individual non-steroidal anti-inflammatory drugs and risk of acute kidney injury: a systematic review and meta-analysis of observational studies. European journal of internal medicine, 26(4), 285-291.
van der Heijden, P. G., van Puijenbroek, E. P., van Buuren, S., & van der Hofstede, J. W. (2002). On the assessment of adverse drug reactions from spontaneous reporting systems: the influence of under‐reporting on odds ratios. Statistics in medicine, 21(14), 2027-2044.
Wen, Y.-K. (2009). Impact of acute kidney injury on metformin-associated lactic acidosis. International urology and nephrology, 41(4), 967.
Wilmer, A., Louie, K., Dodek, P., Wong, H., & Ayas, N. (2010). Incidence of medication errors and adverse drug events in the ICU: a systematic review. BMJ Quality & Safety, qshc. 2008.030783.
Yue, K., Zou, B., Wang, L., Li, X., Zeng, M., & Wei, F. (2017). Prediction of Drug-Drug Interactions Based on Multi-layer Feature Selection and Data Balance. Chinese Journal of Electronics, 26(3), 585-590.
Zhang, S., Zhang, L., Qiu, K., Lu, Y., & Cai, B. (2015). Variable selection in logistic regression model. Chinese Journal of Electronics, 24(4), 813-817.
Zhou, Z.-H., & Feng, J. (2017). Deep forest: Towards an alternative to deep neural networks. arXiv preprint arXiv:1702.08835.