運用決策樹演算法於肝硬化重症病人死亡預測

USING DECISION TREE C4.5 ALGORITHM TO PREDICT THE DEATH OF PATIENTS WITH CIRRHOSIS

10.6338/JDA.202008_15(4).0001

馬瑞菊(Jui-Chu Ma)；林佩璇(Pei-Xuan Lin)；林俊男(Chun-Nan Lin)；鄭婉如(Wan-Ju Cheng)；李佳欣(Chin-Hsin Li)；蕭嘉瑩(Chia-Ying Hsiao)；蘇珉一(Min-I Su)

加護病房 ； 肝硬化 ； 決策樹 ； 死亡預測因子 ； Intensive care unit ； Decision tree ； Cirrhosis ； Predictors of death

Journal of Data Analysis

15卷4期（2020 / 08 / 01）

1 - 14

繁體中文

運用決策樹C4.5演算法，找出影響加護病房肝硬化病人是否死亡之關鍵重要因素，以供相關人員參考。材料與方法：採電子病歷回溯性調查設計，以某區域醫院加護病房2013年8月1日至2015年12月31日肝硬化個案287位進行分析，並以自擬結構性調查表收集資料、以SPSS 22.0分析資料及統計，並使用WEKA3.8.2中的決策樹C4.5演算法產生決策規則以了解哪些因素的組合將可預測入住加護病房之肝硬化病人是否死亡。結果：透過配合不同參數之設定與交叉驗證，求得羅吉斯迴歸預測準確度為94.4%，而決策樹C4.5演算法求得預測的準確度為96.5%，並顯示如果病患「APACHE II score ＞28」且「Albumin ＜=3.2」其死亡機率較高，其他規則尚與APACHE II score 、MELD score、Gender、Viral hepatitis、是否已戒酒、Cr、年齡、Bilirubin等有關。結論：決策樹C4.5演算法較羅吉斯迴歸具有較高的正確率且可產生決策規則更適用於探究加護病房肝硬化病人是否死亡之預測。研究結果將有助醫療人員提早掌握肝硬化病人病情進程並提供病人及家屬合宜之治療方針。

Purpose: Decision tree C4.5 algorithm is used to find out the key factors that affect the death of patients with cirrhosis in the intensive care unit for reference. Material and Method: A retrospective investigation is designed with electronic medical records to make an analysis on 287 patients with cirrhosis in the intensive care unit of a regional hospital from August 1, 2013 to December 31, 2015. A self-designed structural survey is made for data collection, and SPSS 22.0 is used for data analysis and statistics. Decision tree C4.5 algorithm of WEKA3.8.2 is used to generate decision rules and understand which combination of the factors will predict the death of the patients with cirrhosis in the intensive unit. Result: By operating the settings of different parameters and cross-validation, the accuracy of Logistic regression prediction is 94.4% but the accuracy of decision tree C4.5 algorithm is 96.5%. It is also shown that the death of rate for patients is higher when their APACHE II score ＞28 and Albumin ＜=3.2. Regarding other rules, APACHE II score, MELD score, gender, viral hepatitis, quitting alcohol, Cr, age and bilirubin should be considered. Conclusion: Decision tree C4.5 has higher accuracy than Logistic repression, and it can produce decision rule which is more suitable to predict whether patients with cirrhosis in the intensive unit are dead. The result of this study will help medical staff handle the condition of patients with cirrhosis early and then provide sufficient treatment for patients and their families.

1. 張語恬,朱基銘,簡戊鑑,周雨青,楊燦,盧瑜芬,羅慶徽(2007)。比較三種資料探勘演算法預測子宮頸癌五年存活的外部通用性效能。臺灣家庭醫學雜誌，17(4)，222-238。
   連結：
2. Bao, Q.,Wang, B.,Yu, L.,Weng, H.,Ge, J.,Li, L.(2016).A modified prognostic score for critically ill patients with cirrhosis: An observational study.Journal of Gastroenterlogy and Hepatology,31(2),450-458.
3. Berry, M.J.A.,Linoff, G.(1997).Data Mining Technique for Marketing, Sale and Customer Support.Wiley Computer.
4. Demsar, J.,Zupan, B.,Aoki, N.,Wall, M. J.,Granchi, T. H.,Robert Beck, J.(2001).Feature mining and predictive model construction from severe trauma patient's data.International Journal of Medical Informatics,63(1-2),41-50.
5. Escorsell, A.,Mas, A.,de la Mata, M.(2007).Acute liver failure in Spain: Analysis of 267cases.Liver Transplantation,13(10),1389-1395.
6. Hssina B, Merbouha A,Ezzikouri H, Erritali M.(2014).A comparative study of decision tree ID3 and C4. 5.Int J Adv Comput Sci Appl,4,13-19.
7. Karvellas, C. J.,Bagshaw, S. M.(2014).Advance in management and 14. prognostication in critically ill cirrhotic patients.Current Opinion Critical Care,20(2),201-207.
8. Kumar, R.,Shalimar,Bhatia, V(2010).Antituberculosis therapy-induced acute liver failure: Magnitude, profile, prognosis, and predictors of outcome.Hepatology,51(5),1665-1674.
9. Osei-Bryson, K. M.(2004).Evaluation of decisiontrees: A multi-criteria approach.Computers and Operations Research,31,1933-1945.
10. Polat, K.,Güneş, S.(2009).A novel hybrid intelligent method based on C4. 5 decision tree classifier and one-against-all approach for multi-class classification problems.Expert Syst Appl,36,1587-1592.
11. Quinlan, J. R.(1993).C4.5: Programs for Machine Learning.San Francisco, CA:Morgan Kaufmann Publishers.
12. Safavian, S. R.,Landgrebe, D.(1991).A Survey of decision tree classifier Methodology.IEEE Transactions on Systems Man and Cybernetics,21(3),660-674.
13. Tan, P. N.,Steinbach, M.,Kumar, V.(2005).Introduction to Data Mining.Indiana:Addison-Wesley.
14. Warren, J. B.,Yoder, L. H.,Young-McCaughan, S.(1999).Development of a decision tree for support surfaces: A tool for nursing.Medsurg Nursing,8,239-245.
15. William, M. J.(1994).Acute liver failure.The New England Journal of Medicine,329,1862-1872.
16. 李秀蘭,高碧月,林詩偉(2015)。探究生命末期病患與皮膚壓瘡之相關性分析。醫療資訊雜誌，24(2)，11-25。
17. 李育志,阮國彰,鄭昌錡(2015)。肝衰竭病患的血液淨化系統。腎臟與透析雜誌，27(4)，197-200。
18. 翁政雄,洪令莊,呂培豪(2013)。應用決策樹於心臟病預測之研究。第19屆資訊管理暨實務研討會，臺中:
19. 高鴻文,林詩偉,萬書言(2015)。運用決策樹演算法於護理人員離職預測-以某公立醫院為例。醫療資訊雜誌，21(4)，15-29。
20. 陳蓉蓉,洪昌億,陳春賢,鍾青萍,馬成珉,林建雄,張禾坤(2006)。策樹於中西醫腦中風診斷指標結合之應用。醫療資訊雜誌，15(2)，1-15。
21. 衛生福利部（2015，7月）。肝病有藥醫，護肝者聯盟，出動！快找肝膽腸胃科治療．取自http://www.mohw.gov.tw/news/531450997.。