運用資料探勘技術由健康檢查與生活習慣資料建立冠狀動脈心臟病預測模型

Using Technique of Data Mining on Health Examination and Personal Habits for Building the Predictive Models of Coronary Heart Disease

段名聰(Ming-Tsung Tuan)

指導教授 ： 陳澤生

國立成功大學/工學院/工程管理碩士在職專班/碩士(2020年)

摘要

全球多數先進國家早已邁入高齡化社會，更有研究指出在未來二十年內，即將出現超高齡社會，人口老化儼然成為國家亟需研擬對策的社會問題，長照及醫療成為政府必須重視且發展的因應措施。潛藏在人口老化內最棘手的則是慢性病。多數慢性病患者需要耗費相當多時間及金錢在長照及門診上，因此醫院常有人滿為患的情形，如何減輕醫療負擔及提倡健康意識非常重要。科技日新月異的進步，便利了人類的生活，人工智慧的功能逐漸強大，應用領域也越來越廣泛，本研究將運用資料探勘的技術來改善上述的問題，期望能減輕醫院的負擔，並且可以提高民眾的健康觀念。 冠狀動脈心臟病是台灣常見的慢性病，要確認是否罹病需要較精密的檢查以及昂貴的費用，如果可用較為簡易的方式預測，將能替病患省下時間及金錢，並且可以提供醫師輔助診斷，也避免醫療資源浪費。本研究運用資料探勘的技術，從大量健康檢查資料中以決策樹及類神經網路演算法各建立一個疾病預測模型，並且比較預測準確率來提供醫師評估病患是否罹患冠狀動脈心臟病。資料探勘的精神在於從眾多資料內找出潛在的實用資訊，本研究收集最基本的健康檢查報告及生活習慣調查資料，利用R Studio軟體來進行資料探勘，建立冠狀動脈心臟病的預測模型。基本的健康檢查項目雖然不多，但由於本研究運用監督式學習的演算法，在模型建立之時，採用的是已知資料，因此在預測時同樣也只需要輸入相對應的檢查項目，模型就可以即時提供預測結果作為輔助診斷，醫師可以根據健康檢查報告給予受檢人相關建議及健康諮詢，並且讓民眾能更加注意自身的健康情形。

關鍵字

Data Mining ； Coronary Heart Disease ； Decision Tree ； Neural Network

並列摘要

Recently, aging society has impacted most advanced countries in the world. Some researchers indicated that super-aging society will show up in recent twenty years. Consequently, the government must investigate and plan for aging population and put emphasis on the Long-term care and medical services. Chronic disease is the most troublesome problem hiding behind the aging society since aging is not only an immediate personal issue but also a silent factor in crucial public policies, such as pensions, health and long-term care. Most patients with chronic diseases need to spend a lot of time and money on long-term care, along with overcrowded hospital. Consequently, reducing the burden of medical institutions and do the health promotion are quite important. With the progress of the technology, life is more convenient than before. The functions of artificial technology are widely concerned. In this research, we tried to solve the problems above by using data mining technology, anticipating to releasing the burden of hospital, and make people paying more attention on their health. Coronary heart disease (CHD) is a common chronic disease in Taiwan. People always spend much money doing medical examination to check for the disease. If we could predict whether a patient has fallen sick with the coronary heart disease by the proposed simple method, patients can save so much money and time. Moreover, we can decrease the waste of medical resource. Also, the prediction will be supportive diagnosis for doctors to analyze the patient’s healthy examination report. The data mining technique via decision tree and neural network algorithm are used in this research. We established two predictive models based on a great quantity of healthy examinations. We compared the accuracy of the two models, and try to provide information for doctors to evaluate whether patients suffered the coronary heart disease or not. By the detecting and searching skills of data mining, people could discover the needed and potential information from huge database. This research collected a lot of data about the fundamental healthy examination and living habits, and executed data mining to build predictive model of coronary heart disease by R studio software. We used the supervised learning algorithm to build models based on definite data in this research. Because we could only include limited items in the fundamental healthy examination, we input the same categories of data to get the prediction. According to the prediction and general healthy examination report, doctor could provide advice and medical consultation, and remind patients that they should concern much more about their healthy state.

參考文獻

中文部份

1. 王宗道（2002）。血脂肪異常和冠狀動脈心臟病的分子生物學、臨床醫學、及流行病學研究。國立臺灣大學臨床醫學研究所博士論文，台北市，台灣。

2. 王宗道(2001)。心臟血管疾病的預防與治療-淺談心肌梗塞。台大醫網，7(9)，頁26-27。

3. 李師毅(2011)。冠狀動脈心臟病檢測及風險評估雲端服務系統建立之研究。輔仁大學資訊管理學系碩士論文，新北市，台灣。

4. 李俊宏與古清仁(2010)。類神經網路與資料探勘技術在醫療診斷之應用研究。工程科技與教育學刊，7(1)，頁154-169。