請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/20468
標題: | 機器學習之應用: 台灣高科技產業族群非酒精性脂肪肝病之分類與其危險因子之探討 Machine Learning Application: Classification of Non-Alcohol Fatty Liver Disease and its Risk Factors in Taiwanese High-Tech Industry Workers |
作者: | Yu-Han Cheng 鄭羽涵 |
指導教授: | 周呈霙(Cheng-Ying Chou) |
關鍵字: | 非酒精性脂肪肝病,機器學習,分類樹,支持向量機, NAFLD,machine learning,classification tree,support vector machine, |
出版年 : | 2017 |
學位: | 碩士 |
摘要: | 近年來隨著肥胖(症)以及代謝症候群的盛行,非酒精性脂肪肝病儼然成為國人的健康議題,值得注意的是肝癌在台灣十大癌症死因中排名第二,而非酒精性脂肪肝病就有可能導致肝癌。本研究的目標為利用分類樹(classification tree)判定非酒精性脂肪肝病之潛在危險因子,再藉由數種不同的機器學習方法建立學習系統去做分類,同時,我也會比較各個分類器(classifier)的表現,其中分類器包含最近鄰居法(k-nearest neighbor)、引導聚集法(bootstrap aggregating)、隨機森林(random forest),以及支持向量機(support vector machine)。從研究結果可以發現,男性的非酒精性脂肪肝病危險因子可能包含: 代謝症候群、身體質量指數、三酸甘油脂、總膽固醇、年齡、腰臀比、高密度脂蛋白,以及低密度脂蛋白,而在分類器的評估上,支持向量機的表現最好,準確度、敏感度,與特異度分別可達86.9%、90.0%,與81.0%。由此可以推論:將分類樹與支持向量機這兩種方法做結合,對於將男性是否為非酒精性脂肪肝病患者做正確分類是有發展潛力的。另外,此研究將提高國人對於定期健康檢查之重要性的意識,進而預防代謝相關疾病的發生,且在未來針對減少台灣非酒精性脂肪肝病病例的臨床決策必定有所助益。 The prevalence of obesity and metabolic syndrome has led non-alcoholic fatty liver disease (NAFLD) to become a serious health concern during recent years. NAFLD may also lead to hepatoma, which has high mortality rate in Taiwan. The objective of the study is to identify the potential factors of NAFLD by a classification tree (CT) first and then apply machine learning methods to the health examination data to construct the learning system. The performance of several methods including the k-nearest neighbor (KNN), bootstrap aggregating (Bagging), random forest (RF), and support vector machine (SVM) will be compared in this work. I observe that metabolic syndrome, body mass index, triglyceride, total cholesterol, age, waist-to-hip ratio, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol may be the risk factors of NAFLD for males and the SVM classifier gave the best performance (86.9% accuracy, 90.0% sensitivity, and 81.0% specificity). I infer from the study that a combination of decision trees and SVM have the potential to classify NAFLD in males properly. This work can bring more awareness to the importance of regular health checkups to prevent metabolic diseases and aid in the clinical decision making for decreasing NAFLD in Taiwan in the future. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/20468 |
DOI: | 10.6342/NTU201703346 |
全文授權: | 未授權 |
顯示於系所單位: | 統計碩士學位學程 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-106-1.pdf 目前未授權公開取用 | 1.27 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。