臺灣博碩士論文加值系統

目標：為了瞭解身體健康狀況，患者需要進行侵入式檢查，如血液檢查和生化檢查，且許多研究表明空氣污染對人體有害。本研究使用受測者之非侵入式健檢資料以及空氣品質開放資料，建立預測模型。應用此模型在抽血檢查之前，進行肝功能的評估，可以減少醫療成本。
方法：本研究整合37579筆受測者的非侵入式健檢資料，包括基本資料、體格檢查、尿液檢查、問診資料、身體外部檢查和問卷檢查等，以及追蹤1996年至2010年，共15年長期暴露之受測者居住環境空氣品質開放資料。並使用成本敏感學習法以及抽樣技術，分別建立不同的分類器模型來預測目標。預測目標為GPT。
結果：使用成本敏感學習法並設定適當的參數，預測GPT能達到Recall為71.9%，ROC為63.5%。
結論：本研究建立了預測模型，應用此模型在檢查肝功能，可以不必透過抽血與生化檢查，便提早一步發現身體健康狀況是否正常，得以及早發現疾病並接受治療。
關鍵字：非侵入式檢查、健檢資料、空氣品質、開放資料、肝功能、預測模型

Objectives
In order to know the patients' health status, patients need to do invasive tests, such as blood tests and biochemical tests, and many studies have shown that air pollution is harmful to humans. This study uses the non-invasive health information of the subjects and the open data of air quality to establish a prediction model. Apply this model in the blood test before the assessment of liver function, can reduce medical costs.
Method
This study integrates 37579 non-invasive health data, including basic information, physical test, urine test, interrogation data, body check and questionnaire test, and the air quality open data of the living environment for those exposed to long-term exposure from 1996 to 2010. And use the cost-sensitive method and sampling technique, to build a different classifier model to predict the target. Prediction target is GPT.
Result
Using the cost-sensitive method and setting the appropriate parameters to predict Alb, can achieve recall as 71.9%, ROC was 63.5%.
Conclusion
This study builds a prediction model that can be used to detect liver function. It is able to check whether the health status is normal, and early detection of disease for early treatment.
keywords: non-invasive test, health examination data, air quality, open data, liver function, prediction model

目錄
摘要 I
Abstract II
目錄 III
圖目錄 IV
表目錄 V
第壹章 緒論 1
第一節 研究背景與動機 1
第二節 研究目的 2
第貳章 文獻回顧 3
第一節 空氣汙染對人體的危害 3
第二節 成本敏感學習法（Cost Sensitive Learning） 4
第三節 隨機森林（Random Forest） 4
第參章 研究方法 5
第一節 研究流程 5
第二節 研究材料 5
第三節 資料前處理 10
第四節 重新取樣（Resample） 17
第五節 隨機森林（Random Forest） 18
第六節 評估測量（Evaluation Measurement） 19
第七節 敏感學習法 21
第肆章 結果與討論 23
第伍章 結論與建議 24
參考文獻 25
附錄一 健檢資料檢測詳細項目 27
 
圖目錄
Figure 1：研究流程圖 5
Figure 2：標準化的機率範圍量尺 12
Figure 3：測站距離計算示意圖 14
Figure 4：開放資料處理流程圖 13
Figure 5：隨機森林示意圖 18
 
表目錄
Table 1：空氣品質開放資料範例（以「中山測站」2010/01/01的資料整理） 9
Table 2：健檢資料項目統計表 11
Table 3：空氣品質指標 14
Table 4：空氣品質特徵屬性計數器，以中山區測站為例 16
Table 5：混淆矩陣 19
Table 6：代價矩陣的例子 22
Table 7：實驗一，使用非侵入檢測資料建立模型的效果 23
Table 8：實驗二，加入開放資料為屬性的效果 23

[1] 林錫順，「肝病總體檢預防與治療（彩圖版）」，智惠文化，2016
[2] 野村喜重郎，「肝病Q&A」，世茂出版有限公司，2007
[3] 衛生福利部，「105年度死因統計」，衛生福利部統計處，2017，[Online]. Available: http://dep.mohw.gov.tw/DOS/cp-3352-33579-113.html
[4] 楊培銘，「肝與肝病」，1999
[5] Environmental Protection Administration Executive Yuan R.O.C. (Taiwan). (2017, October). Taiwan air quality monitoring network [Online]. Available: https://taqm. epa.gov.tw/taqm/tw/default.aspx
[6] G. Pershagen, "Air pollution and cancer," IARC Scientific Publications, no. 104, pp. 240–251, January 1989.
[7] W. C. Pan, C. D. Wu, M. J. Chen, Y. T. Huang, C. J. Chen, H. J. Su, and H. I. Yang, "Fine particle pollution, alanine transaminase, and liver cancer: A Taiwanese prospective cohort study (REVEAL-HBV)," Journal of the National Cancer Institute, vol. 108, no. 3, March 2016.
[8] O. Raaschou-Nielsen, Z. J. Andersen, R. Beelen, E. Samoli, M. Stafoggia, G. Weinmayr, B. Hoffmann, P. Fischer, M. J. Nieuwenhuijsen, and B. Brunekreef, "Air pollution and lung cancer incidence in 17 European cohorts: prospective analyses from the European Study of Cohorts for Air Pollution Effects (ESCAPE)," The Lancet Oncology, vol. 14, no. 9, pp. 813–822, July 2013.


[9] H. Omidvarborna, A. Kumar, and D. S. Kim, "Recent studies on soot modeling for diesel combustion," Renewable and Sustainable Energy Reviews, vol. 48, pp. 635–647, August 2015.
[10] P. N. Tan, M. Steinbach, and V. Kumar, "Classification: alternative techniques," in Introduction to data mining, 2005, pp. 227–305.
[11] C.X. Ling and V.S. Sheng, "Cost-Sensitive Learning and the Class Imbalance Problem" , 2011, pp. 231-235.
[12] T. K. Ho, "Random decision forests," in Proceedings of 3rd International Conference on Document Analysis and Recognition, August 1995, vol. 1, pp. 278–282. doi: 10.1109/ICDAR.1995.598994
[13] W. Wyman, "Calculate probability of a range using Z Score," 2015. [Online]. Available: http://access-excel.tips/probability-range-z-score/.
[14] TIBCO, "Random Forest Template for TIBCO Spotfire," TIBCO-Community, 2016. [Online]. Available: https://community.tibco.com/modules/random-forest-template-tibco-spotfirer.
[15] R. Kohavi, "A study of cross-validation and bootstrap for accuracy estimation and model selection," LJCAI, Stanford, CA., vol. 14, no. 2, pp. 1137-1145, 1995.
[16] S. Akansha, "Creating, Validating and Pruning Decision Tree in R," edureka, 2015. [Online]. Available: https://www.edureka.co/blog/implementation-of-decision-tree/.
[17]Department of Health Taipei City Government. (2012, April). The short portable mental status questionnaire [Online]. Available: http://www.xyhc.gov.taipei/ct.asp? xItem=1169484&ctNode=30044&mp=109101