以預測市場理論建構傳染病預測模式

Abstract

本計畫有兩個目標。第一是建立一個基於醫療專業人員投入的資訊的互動系統以預測台灣的傳染病疫情。第二是藉由比較這個系統的預測表現和傳統流行病學預測方法（使用同期之期望值做為當期的預測）以評估其優缺點。 為了達成這兩個目標，本計畫實行「以預測市場進行預測」的研究方法。本計畫預計在網路建立上一個的「傳染病預測市場」並邀請醫療專業人員到這個市場上交易「關於台灣疫情的合約」。因為參與的醫療專業人員可以隨時進行交易，所以這些「合約」的價格可以作為反映了最新前線疫情資訊的預測。除了這個即時更新的特色之外，「傳染病預測市場」還提供兩個相較於潛在的其它預測方法的優勢。首先，因為所有參與者可以觀察到所有其他參與者的投入並回饋新的投入到系統之中，所以「傳染病預測市場」是一個互動預測系統。其次，這個市場提供參與的醫療專業人員較佳的誘因──一個參與者的交易如果對價格的預測準確性提升度的貢獻愈高，就可以得到愈多的獎勵。 本計畫的「傳染病預測市場」所交易的合約將包含腸病毒和類流感的高峰週別以及登革熱、腸病毒感染併發重症、流感併發重症的發生首例、疾病高峰週別、流行結束時間和流行程度。在計畫執行中，即時價格將提供給疾病管制局參考。期末報告將針對預測市場在未來疾病管制的應用提出評估與建議。

This project has two objectives. The first is to establish an interactive system to forecast epidemic situation of infectious diseases in Taiwan based on information inputs from medical professionals. The second is to evaluate the strengths and weaknesses of such a forecasting system by comparing its forecasting performance with that of the traditional forecasting methods of epidemiology (using the expected values of the same period as the forecasts for the current period). To achieve these objectives, this project implements the methodology of “forecasting with prediction markets.” The plan of the project is to set up a web-based “prediction market for infectious diseases” and to invite medical professionals as traders of claims about future epidemic situation in Taiwan. Since the participating medical professionals can make trades at any time, the prices of such claims can serve as forecasts reflecting the most updated front-line epidemic information. In addition to this feature of real-time updating, the “prediction market for infectious diseases” provides two additional advantages over potential alternative forecasting methods. First, the prediction market is an interactive forecasting system – all participants can observe the inputs of other participants and feedback new inputs to the system. Second, this market offers better incentives for participating medical professionals to feedback to the system – a participant will get more reward if his or her trading contributes more to the improvement in the forecasting accuracy of the prices. The claims to be traded in the “prediction market for infectious diseases” of this project will cover the peak week of Enterovirus and Influenza-Like Illness and the first occurrence, the peak week, and the ending of Dengue Fever, Enteroviruses Infection with Severe Complications, and Severe Complicated Influenza Cases. The real time prices will be provided to the Center for Disease Control during the execution of the project. The final report will make evaluation and recommendations on future application of prediction markets in disease control.