交通流量對於系統管理者是一個相當重要的依據，本研究嘗試將函數資料分析方法應用到運量預測。其主要研究框架函數型混合預測模型可分成三部分: (1) 函數型資料分群； (2) 函數資料隸屬度分類； (3) 函數型簡單迴歸模型； (4) 混合預測模型。 本研究使用單一路線之捷運站點進出人數作為分析資料，經過資料清洗後共計363天(從2017年4月至2018年3月)。 其結果顯示最好的預測時區在以14個已知時點(τ=14)，預測之CMAPE為12.68%，可提供經營者作為人力指派或是否進行旅客疏導之參考數據。 ;Traffic flow is important for traffic engineers. This study attempts to apply functional data analysis to passenger flows forecasting. The main research framework, the mixture prediction method, can be divided into three parts: (1) functional data clustering; (2) functional data membership classification; (3) functional simple regression model; (4) mixture prediction method. In this study, the number of people entering and leaving the MRT station on a single route was used as analytical data, and the data was cleaned for a total of 363 days (from April 2017 to March 2018). The results show that the best predicted time zone is at 14 known time points (τ = 14) and the predicted CMAPE is 12.68%, it is enough to provide operators to assess whether or not to implement regulatory measures as a reference.
