資料探勘在慢性病預測模式之建構

The Data Mining Approaches to Predict Chronic Diseases

李博智(P. C. Lee)

指導教授 ： 邱昭彰 ； 許光宏

元智大學/資訊學院/資訊管理學系/碩士(2002年)

摘要

本研究主要是使用三維人體體型量測數據當預測變項，與資料探勘研究方法在糖尿病、高血壓與高血脂等慢性病預測模式之建構，在這研究領域是新嘗試。
根據行政院衛生署統計台灣人口十大死因顯示，國人罹患糖尿病、高血壓與高血脂而導致死亡位居台灣十大死因之中。而這三種疾病有些共同特徵都是隨著年齡而快速增加危險因子，與部分相同相關危險因子存在生活環境中。基本上，在國人罹患上述疾病是與許多預知危險因子中的生活型態關係密切。
研究目的是一個預測模式對臨床醫師們平常工作在危險因子無法正常的判斷預測時適時輔助預測危險因子。從預防醫學的觀點而言，一些危險因子的收集可從有效調查代替生化檢驗或是身體檢查。尤其，當預測人體健康狀況時，身體健檢、生活型態變項、家族疾病史等在疾病上扮演重要角色。至於臨床醫師們的觀點，一個有效的預測模式能夠大大的幫助完成診斷、治療與健康教育。在台灣如健康保險轉換成實體系統，對預防醫學角色是更為重要。
資料探勘是利用人工智慧、資料庫與統計等相關技術，從大量資料中挖掘不易發現且有用的資訊或知識，本研究方法使用基因演算法與案例庫推理的混合式資料探勘技術，將權重最佳化與容易探勘擷取出資料庫知識的最佳結合。
研究樣本從2000年七月至2001年七月共收集1370個住院健檢樣本，研究發現三種慢性病的三維預測模式中找到腰臀比是比身體質量指數顯著。在糖尿病方面找到顯著新變項是腰圍輪廓面積。在高血脂方面找到顯著新變項是腰圓周長。在高血壓方面，找到顯著新變項是腰圓周長、身體表面積、左手臂體積等，這都是研究新發現。

關鍵字

高血壓 ； 高血脂 ； 糖尿病 ； 資料探勘 ； 分類 ； 預測

並列摘要

The objective of this paper is to construct a prediction model for chronic diseases such as Diabetes Mellitus, Hypertension, and Hyperlipidemia through the application of methods in Data Mining using three dimensional human body measurements as a new venture of this research filed.
According to records from Department Of Health, Diabetes Mellitus, Hypertension, and Hyperlipidemia were major manifestations among Taiwanese population leading to deaths of top ten causes in Taiwan. These three indications had some characteristics in common as increasing risk with increasing age and sharing the same pool of risk factors in our living environment. Basically, they are such diseases closely related with people’s life-styles as one can predict by some predisposing factors.
The ultimate goal of a prediction model is to foresee risk not normally judged by clinicians’ routine works. From the perspectives of preventive medicine, some risk factors were collected from active survey instead of biochemical tests or physical examinations. Especially, the body measurement, life-style variables, and family history of diseases play important roles in predicting a man’s health. As for clinicians’ points of view, a useful predicting model can greatly help on implementation of diagnosis, treatment, and health education. The role of preventive medicine became more important as health insurance system in Taiwan transforming into prospective payment systems.
The central role of data Mining uses artificial intelligence, database, and statistical methods to extract meaningful information from puzzles of variables and data. This particular study utilizes both genetic algorithm and case base reasoning in hybrid data mining technology. The research suggests this approach to be easy an effective technique to acquire of knowledge from database.
This study has collected 1370 subjects from department of health examination, Chang Gung Memorial Hospital from Jul. 2000 to Jul. 2001 years. Results from predicting selected chronic diseases by anthropometrical and three-dimension measurements are promising and innovative in field of biomedical sciences. Specifically, significant predictors for Hyperlipidemia, Diabetes Mellitus, and Hypertension are wait—hip ratio, waist-profile-area, waist-circum, trunk-surf-area, and left-arm-volume, respectively.

並列關鍵字

Hypertension ； Hyperlipidemia ； Diabetes Mellitus ； Data Mining ； Classification

參考文獻

[1] 王悟師, 林信男：長期飲酒引起的肝硬化及柯沙科夫氏症候群案例, 當代醫學,18卷, 11期, 1991, pp.960-963.

[6] 林榮第, 吳慶南, 李鍾祥：吸菸與健康, 當代醫學, 19卷, 12期, 1992, pp.950-957.

[9] 潘文涵, 黃伯超： 國民營養健康狀況變遷調查, 1993-1996.

[10] Abdolmaleki, P., Buadu, L. D. and Naderimansh, H., “Feature Extraction and Classification of Breast Cancer on Dynamic Magnetic Resonance Imaging Using Artificial Neural Network,” Cancer Letters, Vol. 171, Issue: 2, pp.183-191, October 2001.

[13] Armengol, E., Palaudaries, A. and Plaza, E., “Individual Prognosis of Diabetes Long-term Risks: A CBR Approach,“ Method Inform Med, Vol.40, pp.46-51, 2001.