請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/87579
標題: | 以時間序列分群法結合R/Shiny資料視覺化探討COVID-19傳播 An Investigation to the Spread of COVID-19 via Time Series Clustering and its Data Visualization via R/Shiny App |
作者: | 李奕宏 Yi-Hung Li |
指導教授: | 潘建興 Frederick Kin Hing Phoa |
關鍵字: | 基於模型的遞迴分割法,新型冠狀病毒,視覺化工具,時間序列, model-based recursive partitioning,time series,Shiny,COVID-19, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 隨著2019年新型冠狀病毒的傳播帶來全球性的災害,科學家們開始意識到適合流行疾病分析的機器學習方法論之重要性。而在眾多相關演算法的發展之中,此研究著重在一個適合時間序列資料分析的分群法—「基於模型的遞迴分割法」之時間序列版。在這篇研究中,我快速地探討了幾篇和這個演算法的發展還有一些與本篇發想的方法論有關的文獻,嘗試優化前篇研究所提出之R/Shiny視覺化工具,並且提出了兩大優化這個演算法的方法論,其中包含了基於人口的標準化技巧,以及鄰近域相關特徵的衍生性概念。以台灣一年半的新型冠狀病毒每日確診數資料為研究對象,我不僅在應用此演算法及上述方法論的實驗中得到了一些流行相關的洞見,也透過了統計方法驗證了我的方法論的有效性。完整地分析過實驗後,我針對這個演算法的分群任務和預測任務都個別提出了相對應的建議參數設定。最後,針對開發過程和實驗過程中的限制,我列舉了幾項本研究未來可能可以優化的方向。 The outbreak of COVID-19 raised awareness of the need to develop machine learning methodology suitable for epidemiological pattern recognition. This study focuses on time-series model-based recursive partitioning and one of its derivative applications for visualization. Several literature reviews on related studies including the evolution of this methodology are provided. This study documents the improved functionality enhanced user-friendliness of an R/Shiny application for visualizing time-series model-based recursive partitioning and proposes several methodologies to strengthen the computational results from this algorithm. With Taiwan township-level COVID-19 spread data, several experiments were conducted to gain insights on the spread patterns and validate the effectiveness of this study’s main contributions, which include the population-based scaling technique and the derivative concept of adjacent domain-relevant attributes. Finally, various option configurations for clustering and forecasting tasks based on this epidemic-related algorithm are suggested. After considering the limitation of this work, several promising future directions are provided at the end of this study. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/87579 |
DOI: | 10.6342/NTU202300086 |
全文授權: | 同意授權(全球公開) |
電子全文公開日期: | 2025-01-10 |
顯示於系所單位: | 資料科學學位學程 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-1.pdf 此日期後於網路公開 2025-01-10 | 8.11 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。