應用資料探勘技術於成人健康檢查之慢性病預防

Apply data mining techniquse to adult health examination for prophylaxis of chronic illness

張士瑋(Shih-Wei Jhang)

指導教授 ： 洪士程

朝陽科技大學/資訊學院/資訊工程系/碩士(2014年)

摘要

由於人口老化及人口結構的改變，高齡化的相關社會議題，漸漸受到大家的關注及重視。人口老化帶來醫療費用高漲、慢性病、身體功能退化、殘障等問題成為家庭、社會、國家的沉重負擔。因此，藉由健康檢查來達到早期發現早期治療等目標是根本之道。而成人健康檢查資料中，潛藏著一些未曾探索的重要資訊與知識。資料探勘方法可以做資料類型的發掘與萃取，而對一資料組做最佳的詮釋。本篇碩士論文藉由資料探勘技術，以找出成人健康檢查資料庫中，檢驗數值異常項目的組合。本研究以台北市某銀行所屬機構員工體檢的1000筆資料為資料庫，資料分析分成二大階段：一般統計分析與資料探勘分析。一般統計分析稱為資料的前處理，資料處理後，以供第二階段資料探勘使用。針對檢驗數值具有很多的類型數量且連續性質屬性的分類問題，提出一個結合分群與決策樹的分類器。這個分類器的基本概念是先把所有類型細分成群集，利用關鍵屬性可完成群集的分割，並產生一個群集分割樹。在群集分割樹的終端群集，則套用決策樹的方法，以降低整個分類規則庫的大小並減少計算的複雜性。最後建立出一套疾病的預測模型，透過健康檢查資料屬性去分析異常項目的相關性，並挖掘其中隱含規則，將分析結果建立出模型提供給醫生做為輔助參考，用來提高診斷出慢性病的正確性。

關鍵字

分群 ； 資料探勘 ； 慢性疾病 ； 成人健康檢查 ； 決策樹

並列摘要

Due to increased number of aging population and population structure changes, people are concerned about aging-related problems that put great burdens and make enormous impacts on the family, society, and nation. The problems include increased medical expense, chronic diseases, and degeneration of the elderly. Aging population brings along chronic disease, deterioration of boding function and disabilities resulting in increase medical expenditure and heavy burden to the family, society and the nation as a whole. Regular health examination and early treatment are the most effective solutions to this problem. Important but non-intuitive information and knowledge in the adult health examination database can be found, extracted and organized using data mining. This thesis applied the data mining techniques to enhance understanding of abnormal item combinations in adult health examination data. Researchers collected a total of 1, 000 health examination records of a bank in Taipei for use in this study. Data analysis consisted of the two parts of general statistics and data mining. General statistics was a pre-process used to clarify and organize data in preparation for data mining. A classifier combining clustering and decision tree is proposed to solve the classification problem with large number of classes and continuous attributes. Critical attributes are used to perform the cluster splitting and generate a cluster splitting tree. The decision trees for the terminal clusters in the cluster splitting tree are applied so as to reduce the size of the classification rule set and hence reduce the computational complexity. This research uses the data mining techniques to explore abnormal item combination on community health screening services for the elderly. Besides, related factors of the information and knowledge are also discussed. Suggestions may serve as a useful reference for doctor to improve the correct diagnosis of a chronic illness.

並列關鍵字

data mining ； chronic illness ； adult health examination ； Clustering ； decision tree

參考文獻

[17] N. Bhargava, R. Bhargava, and M. Mathuria, "Decision Tree Analysis on J48 Algorithm for Data Mining," International Journal of Advanced Research in Computer Science and Software Engineering, vol. 3, pp. 1114-1119, June 2013.

[6] X. Chang and J. H. Lilly, "Evolutionary design of afuzzy classifier form data," IEEE Transactions on System,Man and Cybernetics, vol. 34, pp. 1031-1044, April 2004.

[1] M. W. Yu, H. C. Chang, Y. F. Liaw, and S. M. Lin, "Familial risk of hepatocellular carcinoma among chronic hepatitis B carriers and their relatives.," Journal of the national Cancer Institute, vol. 92, July 2000.

[2] S. M. Grundy, "Holesterol and Coronary Heart Disease," Scandinavian Journal of Clinical & laboratory Investigation - Suppllment vol. 199, pp. 17-24, 1990.

[3] T. Denoeux, "A neural network classifier based on Dempster-Shafer theory," IEEE Transactions on System,Man and Cybernetics, vol. 30, pp. 131-150, March 2000.