臺灣博碩士論文加值系統

流行性感冒是由流感病毒而引起的急性呼吸道感染疾病，部分患者可能會引發嚴重的併發症，甚至造成死亡。流感流行期時長不定，通常從12月至隔年1月份進入高峰，也可能延長至隔年2、3月，在人口密集的地區容易發生群聚感染，而臺灣人口主要集中在臺北、新北、桃園、臺中、臺南、高雄六個縣市，相對於其他縣市較為密集，因此當流行性感冒爆發時，大量患者湧入醫療院所，往往造成人力、醫療資源不足。現今大眾可以透過網路獲得疾病相關資料以及資訊，包括疾病的症狀、預防、治療等，大多都是利用輸入關鍵字查詢，而最常使用且方便的查詢工具為谷歌搜尋引擎，特定聲量資訊可反應疾病對民眾的影響，因此可以透過網路聲量數據來監控流感的爆發。目前也有許多研究是探討流感與氣象資料的相關性，以及氣候因子影響流感病毒傳染性的強弱，包括溫度、濕度、降雨量等氣候因子，進而提高對流感的爆發進行預測準確性。而過去研究大多使用機器學習做預測，或者只單獨使用聲量資料或是氣象資料等數據，本研究結合與流感傳染力強弱相關的氣象、空氣品質，以及與疫情嚴重程度有關的聲量資料，使用深度學習中的長短期記憶模型，預測臺灣六都兩週後的類流感發生率，並透過移動流行區間法，將發生率分成不同流行程度的警報級別。本研究定義一聲量關鍵字蒐集流程，確保所蒐集的關鍵字能提升預測的準確度並具有效性，研究結果確定了使用12週的特徵資料所訓練的模型效果較好，且加入聲量特徵可提高預測相關係數並降低誤差，也驗證特徵資料以週為單位相對於以天為單位，更能提升預測類流感發生率的準確度。及早預警流感的爆發以及疫情嚴重程度的分級，讓決策者能有準確的判斷依據，並做出正確的決策，一般民眾也能提早預防，進而減少醫療資源的支出以及因流感造成的死亡。

Influenza is an acute respiratory infection caused by the influenza virus. Some patients may cause serious complications or even death. Flu pandemic duration varies and cluster infections are prone to occur in densely populated areas. Therefore, when a flu outbreak, a large number of patients flow into medical institutions, often resulting in insufficient human and medical resources. The objective of this study is to give an early warning two weeks before the outbreak of influenza, and reduce the expenditure of medical resources and death due to influenza. In this research, that combined with weather, air quality, and network volume data is used as the input features where weather and air quality are related to the strength of influenza infection, and volume is related to the severity of the epidemic. The Long Short-Term Memory model in deep learning is developed and trained to predict the incidence of influenza-like illness rates located in major cities in Taiwan for two weeks later. The level of the epidemic in this study is based on the Moving Epidemic Method. This study also defines a comprehensive process to collect the network volume of keywords to ensure that it can improve the accuracy and effectiveness of prediction. The results of the study show that (1) the feature data divided into 12-weeks cycle length can get better performance in the trained model; (2) the network volume features can improve the prediction correlation coefficient and reduce the error; and (3) the feature data in units of weeks is better than in days from the accuracy of predicting the incidence rate of influenza. Early warning of the outbreak of influenza and the classification of the severity of the epidemic allows decision-makers to have accurate judgments and make correct decisions.

第一章 緒論 1
1.1 研究動機與背景 1
1.2 研究目的 1
第二章 文獻探討 3
2.1 流感相關背景知識 3
2.1.1 疾病介紹 3
2.1.2 感冒、流感與類流感比較 3
2.1.3 預防方式 4
2.2 影響流感傳播之因子 5
2.3 聲量用於醫療之監控 6
2.4 類流感之預測 9
第三章 研究方法 11
3.1 研究資料 11
3.1.1 氣象資料 11
3.1.2 空氣品質資料 17
3.1.3 聲量資料 18
3.1.4 類流感資料 22
3.2 長短期記憶模型 23
3.3 預測方法 25
3.3.1 深度學習模型 25
3.3.2 移動流行區間法 26
3.4 模型評估 29
第四章 實驗 31
4.1 實驗一:特徵資料週數 31
4.2 實驗二:特徵資料以天、以週為單位 32
4.3 實驗三:聲量資料是否幫助預測 34
4.4 實驗四:深度學習與機器學習比較 41
第五章 討論 45
5.1 長短期記憶模型序列資料的長度 45
5.2 資料的精細度(granularity) 45
5.3 聲量資料的重要性 45
5.4 深度學習的優勢 45
第六章 結論 47
參考文獻 48

[1] 衛生福利部疾病管制署, "台灣流感速訊," Mar. 2019. [Online]. Available: https://www.cdc.gov.tw/File/Get/02RDONATv6PYQMim5j6oGA.
[2] A. C. Nage, M. H. Tsou, B. H. Spitzberg, L. An, J. M. Gawron, D. K. Gupta, . J. A. Yang, S. Han, K. M. Peddecord, S. Lindsay and M. H. Sawyer, "The Complex Relationship Of Realspace Events And Messages In Cyberspace: Case Study Of Influenza And Pertussis Using Tweets," Journal of Medical Internet Research, p. 15(10):e237, Aug. 2013.
[3] Y. Tausczik, K. Faasse, J. W. Pennebaker and K. J. Petrie, "Public Anxiety and Information Seeking Following the H1N1 Outbreak: Blogs, Newspaper Articles, and Wikipedia Visits," Health Communication, p. 179–185, Aug. 2011.
[4] F. Saberian, A. Zamani, M. M. Gooya, P. Hemmati, M. A. Shoorehdeli and M. Teshnehlab, "Prediction Of Seasonal Influenza Epidemics In Tehran Using Artificial Neural Networks," in 2014 22nd Iranian Conference on Electrical Engineering (ICEE), Tehran, Iran, 2014.
[5] 衛生福利部疾病管制署, "流感併發重症疾病介紹," Jan. 2020. [Online]. Available: https://www.cdc.gov.tw/Category/Page/HMC9qDI4FA-gDrbcnFlXgg.
[6] 編輯部, "感冒、流感、類流感有什麼不同？," 康健雜誌, no. 30, 14 9 2012.
[7] 衛生福利部疾病管制署, "流感併發重症傳染病防治工作手冊," Jan. 2020. [Online]. Available: https://www.cdc.gov.tw/Category/MPage/6YQ32GG9EOdXNTjPqUDT1A.
[8] medpartner, "流行性感冒症狀有哪些？如何預防及治療？圖文完整解說！," 美的好朋友, Dec. 2019. [Online]. Available: https://www.medpartner.club/influenza-prevention-treatment/.
[9] A. C. Lowen, S. Mubareka, J. Steel and P. Palese, "Influenza Virus Transmission Is Dependent on Relative Humidity and Temperature," PLOS ONE, p. 3(10): e151, Aug. 2007.
[10] D. Lopez, M. Gunasekaran, B. S. Murugan, H. Kaur and K. M. Abbas, "Spatial Big Data Analytics of Influenza Epidemic in Vellore, India," in IEEE International Conference on Big Data (Big Data), Washington, DC, USA, 2014.
[11] Z. Xu, W. Hu, G. Williams, A. C. Clements , H. Kan and S. Tong, "Air Pollution, Temperature And Pediatric Influenza In Brisbane, Australia," Environment International, vol. 59, pp. 384-388, Sep. 2013.
[12] X. Zhou, Q. Li, Z. Zhu, H. Zhao, H. Tang and Y. Feng, "Monitoring Epidemic Alert Levels By Analyzing Internet Search Volume," IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, pp. 446-452, Feb. 2013.
[13] A. F. W. Ho, B. Z. Y. S. To, J. M. Koh and K. H. Cheong, "Forecasting Hospital Emergency Department Patient Volume Using Internet Search Data," IEEE Access, pp. 93387-93395, Jul. 2019.
[14] 周書任，張昭憲,“以模型融合配合社群網路資料進行流感趨勢預測,” Journal of Information, Technology and Society, Dec. 2018.
[15] 交通部中央氣象局, "觀測資料查詢(CWB Observation Data Inquire System)," [Online]. Available: http://e-service.cwb.gov.tw/HistoryDataQuery/index.jsp.
[16] 衛生福利部疾病管制署, "季節性流感防治," Jan. 2020. [Online]. Available: https://www.cdc.gov.tw/Category/QAPage/DQWXG19u2cXMH1jwGKXHug.
[17] 衛生福利部疾病管制署, "季節性流感防治工作手冊," Dec. 2018. [Online]. Available: https://www.cdc.gov.tw/uploads/files/b806ae73-d2f8-403a-a754-d698f6ea4cce.pdf.
[18] 衛生署疾病管制局, "全球流感的流行病學," Mar. 93. [Online]. Available: https://www.cdc.gov.tw/File/Get/uNzmh1MVBm-xt93Igw6djA.
[19] F. Gers, J. Schmidhuber and F. Cummins, "Learning to Forget: Continual Prediction with LSTM," in 1999 Ninth International Conference on Artificial Neural Networks ICANN 99, 1999.
[20] 衛生福利部疾病管制署, "疫苗接種計畫篇," Sep. 2019. [Online]. Available: https://www.cdc.gov.tw/Category/QAPage/jWgjO_d826X_F9TURP2_Qg.
[21] W. H. O. (WHO), "Global Epidemiological Surveillance Standards for Influenza," Jun. 2014. [Online]. Available: https://www.who.int/influenza/resources/documents/WHO_Epidemiological_Influenza_Surveillance_Standards_2014.pdf?ua=1.
[22] T. Vega, J. E. Lozano, T. Meerhoff, R. Snacken, J. Mott, R. O. d. Lejarazu and B. Nunes, "Influenza Surveillance In Europe: Establishing Epidemic Thresholds By The Moving Epidemic Method," Influenza Other Respir Viruses, pp. 7(4):546-58, Jun. 2012.