環顧國內整體運輸需求模式,其中歷史最久之臺北模式演進至第四代模型,皆以旅次基礎模型構建。由於旅次為活動參與下之衍生物,傳統旅次基礎模型未能從活動參與之觀點探討,具有理論上之缺陷。國外於近二十年來,致力發展活動基礎之整體需求模型推估旅次鏈並應用至規劃實務中,已有合理且有效的推估結果。 時逢臺北市政府捷運工程局辦理「臺北都會區整體運輸需求預測模式建立-旅次行為調查及旅次發生模組」之「家戶旅次活動特性訪問調查」甫完成,具有大規模之調查樣本與穩定之調查品質,有助於瞭解臺北都會區之在地旅運行為,進而構建相關模型。 以活動理論為基礎之整體運輸需求模型整體架構相當龐大,本研究提出可能之整體概念流程,並先聚焦於家戶活動產生及成員指派模型,因應模型探討內容之特殊性分別應用階層線性模型與羅吉特模型構建。本研究獲得之重要成果之一,在於檢視旅運資料時,顯示其具有層級架構,故將變數分層處理,進而析離出變數間之脈絡效果;此外,對於活動指派之程序性建立,本研究確認臺北地區活動之參與特殊性,並用以修正成員指派順序。根據以上之脈絡效果及旅運特殊性,本研究建構之模型參數具有驗證合理性,且相較於傳統模型具有良好的推估結果。
Taipei Rapid Transit Systems Demand Model (TRTS model) has been established for about 30 years and developed for the 4th generation. However, TRTS model is still based on the conventional method - trip-based model. Trip is derived demand from activity participation; therefore, trip-based model is theoretically incomplete. On the contrary, the activity-based travel demand model has been developed and applied in some U.S. metropolitan areas in planning practice with great and reasonable modeling effect. The Department of Rapid Transit Systems (DORTS) of Taipei City Government upgraded the latest TRTS-IV in 2009, which collected 9,000 household travel surveys in Taipei Metropolitan area. The characteristics of local travel behavior could be characterized from the large and stable survey data, which is helpful to establish related models as well. We conceptualized the probable framework of activity-based travel demand model with firstly focusing on the household activity participation and membership assignment models. By applying hierarchical linear model for the former and logit model for the latter, the hierarchical framework of exogenous variables was specified; therefore, variables needed to be separated with different levels to reflect context effect. Furthermore, the sequence of membership assignment was identified based on local travel behavior following hierarchical model. By the context effect and the local travel behavior, the parameters of models were verified and the estimation effect was shown better than conventional models.