流感之巔：探索病毒的盛衰潮流

Abstract

摘要 隨著過去幾年來COVID-19疫情成為國際關注的突發公共衛生事件後，這場規模極大的疫情傳染，讓全球認識到傳染病傳播的危險性和對社會所造成的深遠影響。然而除了COVID-19疫情之外，流感是一種持續存在且具有潛在爆發風險的傳染疾病，是一個更值得探討的公共衛生議題。因此，本文希望透過此專題研究，深入了解臺灣2010年至2019年間，每週類流感健保就診人次的趨勢，以及預測在沒有其他疫情影響之下的流感就診人數變化。本文使用衛生福利部疾病管制署在政府資料開放平台所提供的2010年1月至2019年12月每週類流感健保門診及住院就診人次統計資料，進行時間序列的建模與預測，並考慮時間序列迴歸模型、ARIMA模型及指數平滑法三種不同模型方法，利用 Mean Square Error (MSE)、Root Mean Square Error (RMSE)、Mean Absolute Percent Error (MAPE)、Mean Absolute Error (MAE)和R平方等五個指標來比較三種不同方法的預測效果與解釋能力，以選出最佳的預測模型。最後結果顯示，ARIMA的預測表現最佳，可做為本研究最終預測模型，且從ARIMA模型可發現上一週的類流感就診人數與本週呈負相關，而與年的下一週的類流感就診人數與本週呈正相關。此外，還預測2020年1月至2020年12月在沒有COVID-19發生的情況下每週類流感就診人次分布變化。

Abstract As the COVID-19 epidemic has become a public health emergency of international concern in the past few years, this extremely large-scale epidemic has made the world aware of the dangers of the spread of infectious diseases and its profound impact on society. However, in addition to the COVID-19 epidemic, influenza is an infectious disease that persists and has potential outbreak risks, and is a public health issue worthy of discussion. Therefore, this article hopes to use this special study to gain an in-depth understanding of the trend in the number of weekly influenza-like health insurance visits in Taiwan from 2010 to 2019, and to predict the changes in the number of influenza-like visits without the impact of other epidemics. This article uses the statistical data on weekly influenza-like health insurance outpatient and inpatient visits from January 2010 to December 2019 provided by the Department of Disease Control and Prevention of the Ministry of Health and Welfare on the government information open platform, to conduct time series modeling and prediction, and considers time Three different model methods: sequence regression model, ARIMA model and exponential smoothing method, using Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Percent Error (MAPE), Mean Absolute Error (MAE) and R-squared, etc. Five indicators are used to compare the prediction effects and explanatory capabilities of three different methods to select the best prediction model. The final results show that ARIMA has the best prediction performance and can be used as the final prediction model of this study. From the ARIMA model, it can be found that the number of influenza-like hospitalizations in the previous week is negatively correlated with this week, and is negatively correlated with the number of influenza-like hospitalizations in the next week of the year. The number of people is positively correlated with the week. In addition, changes in the distribution of weekly influenza-like hospital visits from January 2020 to December 2020 in the absence of COVID-19 were also predicted.