| 摘要: | 都會區之基地開發常由於現有交通設施之不足，而對鄰近地區造成嚴重交通衝擊，如何做好事前之規劃與評估，乃為都市與運輸規劃之重要課題；然而，若要基地開發案之詳細的交通衝擊評估(Traffic Impact Assessment, TIA)，則需耗費相當大的調查分析成本，故本研究欲以較簡單之線性迴歸分析方式，針對都會區一般性之中小型基地開發，進行道路交通衝擊預測，研究對象以台北市在民國84 ~ 85年取得使用執照之基地開發案為例。 本研究所建立之道路交通衝擊預測模式乃以線性多元迴歸(Multiple Linear Regression)分析為主軸，在預測變數與準則變數的選取方面，本研究乃參考過去文獻與施行經驗，以及與交通衝擊相關之各項規定，並考量資料蒐集之可行性，共計選定14個預測變數x，大致可分為「基地開發規模」、「基地開發型態」，與開發基地「鄰近地區之社經條件與交通狀況」等分析構面，至於衝擊準則變數y，則以開發基地主要鄰接道路之「旅行速率之變化量」代表，接著並對各變量之資料來源與蒐集方式作適當的定義與說明；本研究針對台北市在民國84 ~ 85年間，取得使用執照之開發基地進行分析，共計蒐集到226個樣本，首先以「集群分析」方式，將樣本依其特性予以分群，共將樣本分為2群，其中「集群1」開發基地的特性與「住宅區」之開發基地較為相近，「集群2」則與「商業區」較為相近，接著藉由「相關分析」檢視變數間之線性相關程度，刪除相關性過高之分析變量，最後則利用逐步迴歸方式，分別對2個群體建立衝擊預測模型，其中「集群1」以開發基地「總樓地板面積」、「容積率」、「小汽車停車位數」與開發基地鄰近地區之「人口密度」、「平均所得」，以及主要鄰接道路之「車道數」、「車流量」與「旅行速率」等8項變量對道路衝擊影響較大，調整後之判定係數達0.5323，而「集群2」則可以開發基地之「總樓地板面積」與「容積率」、開發基地內「商業/辦公用途」之樓地板面積所佔比例，以及主要鄰接道路之「旅行速率」等4變量對道路衝擊進行預測，調整後之判定係數達0.6068。 另外，由於考量到資料蒐集與變數定義時所能提供之資訊並不完整，且多包含許多主觀判斷因素在內，故再以模糊線性迴歸(Fuzzy Linear Regression, FLR)方式進行分析，為要求變數之獨立性，故以先前逐步迴歸之分析結果進行模糊迴歸之分析，本研究以線性規劃(Linear Programming)方式求解(1)「模糊參數總模糊度最小」與(2)「估計值模糊度最小」二種模糊迴歸模型，在模型(1)中，僅鄰近地區之「平均所得」為模糊，其他所有變數皆為明確型，在模型(2)中，則所有變數皆為明確型，僅截距為模糊，顯示衝擊預測系統之解釋能力在此二模型中皆相當理想；而在不同之模糊迴歸配適水準H下，所獲得之結果差異不大，故未來若採用模糊迴歸方式進行預測，則可採用H = 0.0進行分析。 本研究所建立之基地開發道路交通衝擊預測模式，主要乃針對台北市之基地開發案進行分析，未來可應用於其他都會區一般性之基地開發衝擊預測分析，並可搭配鄰近地區之交通現況或其他衝擊評估準則，提供基地開發之使用執照或建築執照等之初步審查或評估的參考依據。 As the short of traffic facilities in urban area usually falls behind the demand of site developments, the traffic impact assessments (TIA) for site developments are becoming important issues in urban and trasportation planning. A detail evaluation and investigation of the traffic impact from a site always costs increditablely high. Meanwhile, the usual sites being developed are in general size, and not each of them is worthy to do that. Consequently, an ordinary traffic impact forecast model should be proposed to deal with the first trial of a developing site. The goal of this research was right to establishing a regural forecasting model to estimate the road traffic impact from common sites. The demonostrated samples are those sites had got the operation licenses during 1995 and 1996 in Taipei. In this research, the multiple linear regression analysis was applied to forecast the road traffic impact. The selections of the independent variables and the dependent variable were refered to the literatures and the relative regulations or experiences. And the practicablilities of data collections were considered, too. 14 independent variables were brought up initially according to the three components: the “site scale”, the “site pattern” and the “soc-economics and traffic situation surrounding the site”. And the only dependent variable was determined to be the difference of velocity in the major adjacent road between the sites being operated and not yet. We got 226 samples, and 2 clusters are suggested standing on the result of the “cluster analysis”. The samples of the 1st cluster were similar to the sites in the residential district while the ones of the 2nd cluster were considered as those in the business district. After inspecting the linear relationships between all the variables by “correlation analysis”, “stepwise regression analysis” was used to build up the road traffic impact forecast model of these 2 clusters. The key factors affecting the traffic impact in the 1st cluster were “the floor area”, “the volume”, “the automobile parking spaces supplied“of the developing site, “the population density”, “the average earnings” in the neighborhood, and “the number of lanes”, “the traffic flow rate”, and “the velocity” in the major adjacent road. The 8 variables could take effect to 0.5323 in this forecasting system. In the meantime, 4 key factors were brought up in the 2nd cluster: “the floor area”, “the volume” of a site, “the rate of the floar area using for business or office purpose”, and “the velocity” in the major adjacent road. The adjuted cofficient of determination could be reached to 0.6068 by using this regression model. Moreover, pondering over the lack of robustness in those analytical variables selection and the data collection procedures, we looked for the fuzzy linear regression (FLR) models for further analyses. 2 common-used linear programming (LP) based models were solved in this research: (1) minimizing the sum of the forecasting parameters’ fuzzy intervals, (2) minimizing the sum of the estimations’ fuzzy intervals. Only the parameter of “the average earnings” in the neighborhood was fuzzy in model (1), all others including the intercept were crisp. And each independent variable in model (2) was crisp, while the intercept was fuzzy. That means the FLR models are toward crisp. Furthermore, different value of H didn’t cause wide variation whether in which one of the 2 FLR models. Once the FLR models were adopted in future days, the lowest value of H = 0.0 would be wonderful in analyses. Though the road traffic impact forecast model was demonostrated by the past Taipei city samples, it can be applied to the site developing decision analysis by cooperating with other aspects of impact evaluation models or with the adjacent traffic situation. |
