資料採礦技術應用於修正版國際攝護腺症狀量表之量度發展研究

Application of Text Mining on Traditional Chinese Medical Diagnosing Acne

林柏村(Bo-Tsun Lin) ； 劉志光(Chih-Kuang Liu)

《Journal of Data Analysis》 9卷2期 (2014/04) Pp. 23-37

https://doi.org/10.6338/JDA.201404_9(2).0003

摘要

攝護腺肥大為中老年男性普遍容易得到的疾病之一，並且隨著年齡的增長，罹患該病症的機率會逐漸升高，一般採用國際攝護腺症狀（International Prostate Symptom Score-IPSS）量表作為攝護腺肥大症狀嚴重程度之評估量表。但該量表設計內容過於繁瑣冗長且內容模糊不清。因此本研究欲透過蒐集良性攝護腺肥大症合併下泌尿道症狀（BPH/LUTS）病人填寫之IPSS量表以及修正版之WINF量表進行資料的分析。透過降低量表題項維度以及縮減答項尺度的方式，簡化目前的症狀評估量表。本研究之主要目的是透過改進修正版的攝護腺症狀量表之答項尺度，並透過資料採礦技術驗證新修正後症狀量表的可行性。透過答項尺度的縮減，並藉由資料採礦技術對選定的四種答項組合，進行建模方法的比較。而四種組合分別是：組合一（時間≦1/5、1/5＜時間≦1/2、1/2 ＜時間≦1）、組合二（時間＜1/2、時間=1/2、時間＞1/2）、組合三（時間=0、0＜時間≦1/5、1/5＜時間≦1）、組合四（時間=0、0＜時間≦1/5、1/5＜時間≦1/2、1/2＜時間≦1）。結果顯示支援向量機（SVM）法比其他建模方法（KNN演算法、邏輯斯迴歸（LogisticRegression)法、神經網路(Neural Network)法及貝氏網路（Bayesian Network）法）更適合建構此量表資料之分類模型，其精準度為63.56%~97.27%。從不同建模方法下比較四種組合，結果顯示組合二的準確率較精準，為82.53%~97.27%。透過方法與組合的交叉比較後，得到最佳組合為SVM建模方法下之組合二，其整體準確率為84.76%至97.27%。經過多次驗證模型所得到之結果顯示模型穩定。最終以組合二做為修正後之WINF量表的答項，並可將其用於替代IPSS量表。

關鍵字

資料採礦

並列摘要

Acne is also called pimples. It most commonly affects the face and the back of shoulder, and it takes the form of tubercles, pustules, or cyst. When the symptom is worse, it may have a scar on the face. The most acne patients are adolescents, and they are more mind their appearance, therefore, they will be affected their physical and mental than others. Traditional Chinese medical can be traced to long time ago. It was ancient's experiences, and be handed down. According to traditional Chinese medical to cure acne, has unique efficacy and effectiveness. Along with the advances in traditional Chinese medical, there are more and more literatures referring to traditional Chinese medical diagnosing Acne. The purpose of the study was to discover the hidden knowledge between these abundant literatures. First, the sentence in the literatures is divided into many independent phrases. Then, this study uses text mining technique to discover the patterns between key phrases. Finally, this study quantifies the results of text mining, and use data mining technique to discover the associations between different symptoms and Chinese herbs, different external therapies and Chinese herbs. The expectation of the study is to provide Chinese medical educators with a greater understanding to diagnosis acne patients.

並列關鍵字

Text Mining ； Data Miming ； Acne ； Traditional Chinese Medical ； International Prostate Symptom Score (IPSS) ； Benign Prostatic Hyperplasia (BPH) ； Lower Urinary Tracts Symptoms (LUTS) ； WINF

參考文獻

https://us.sagepub.com/en-us/nam/reliability-and-validity-assessment/book438