請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/71568
標題: | 以LDA主題模型探討線上醫療社群之癌症病人需求 An exploration on cancer patients needs by LDA topic modeling in online health community |
作者: | Yu-Chun Huang 黃羽均 |
指導教授: | 余俊瑜 |
關鍵字: | 主題模型,LDA,健康需求探勘,非監督式學習,癌症, Topic models,LDA,Healthcare needs,Unsupervised learning,Cancer, |
出版年 : | 2019 |
學位: | 碩士 |
摘要: | 多年來,癌症高居台灣十大死因之首。有效的資訊可以緩解癌症病患的焦慮感,故癌症病患有其資訊需求。由於實體資源往往無法提供足夠資訊,癌症病患轉而使用線上資源以滿足他們的資訊需求。近年來,癌症病患漸漸增加對線上癌症病友社群的使用。
本研究的目標為定義出線上癌症病友社群所討論的主題。因應病患使用線上癌症病友社群的趨勢,有效的分析社群貼文、研究病患所討論的內容,能讓研究人員更了解患者,並提供更好的服務。 本文研究所使用的素材為病患在兩組線上癌症病友社群的貼文。由於社群媒體資料為非結構性資訊,本研究選擇主題模型作為主要的研究方法。在搜尋相關文獻後,本研究選定LDA主題模型對此二文本進行分析。 本研究成功擷取病患在兩個不同社群中所討論的主題,包含臉書社群中的六個主題 ”副作用”、”支持”、”治療”、”新藥”、”問題”``、”祝福”以及批踢踢抗癌版中的五個主題”照護”、”手術”、”病友團體”、”病患狀況”、”藥物”。 本研究證實主題模型能協助研究員定義病患需求,從大量文本裡擷取出隱含的文義。 Background: Cancer patients utilize online information to fulfill their information need. Cancer patients and their relatives increasingly participate in online disease communities. These platforms allow cancer patients to exchange their experience of dealing with the disease by posting open-end discussions. Objective: The aim of this research is to automatically identify different cancer-related topics discussed on online cancer patient community. We use text-mining techniques to examine what patients say about their experiences during cancer journey and to address their unmet needs. By analyzing these user-generated contents, we can get a better understanding of how people participate in the online discussions. Methods: We applied LDA models on the two datasets collected from the Facebook group “Anti-Cancer Alliance” and the “Anti-Cancer” board in PTT, a well-known BBS forum in Taiwan. We conducted some relevant data preprocessing, then we applied LDA model to more than 10,000 discussions to study what cancer patients say online about their disease journey. Results: Experiment results demonstrate that health-related hot topics primarily include 6 topics from posts of Facebook users, which are “Side effect”, “Support”, “Treatment”, “New drug”, “ Questions” and “Blessing”; and 5 topic from BBS users, which are “Care”, “Surgery”, “Support group”, “Patient condition” and” Medicine”. Conclusions: This research shows the potential for extracting keywords to confirm the clinical distinction, and text-mining can help objectively understand the concerns of patients by generalizing their large number of subjective illness experiences. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/71568 |
DOI: | 10.6342/NTU201900229 |
全文授權: | 有償授權 |
顯示於系所單位: | 商學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-108-1.pdf 目前未授權公開取用 | 1.33 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。