公共自行車租借量之影響因素分析-地理加權迴歸和函數資料分析方法之應用

Abstract

近年隨著都市成長，運輸需求大幅增加，同時也產生了二氧化碳排放量增加、溫室效應等現象，各國政府開始鼓勵民眾步行、騎乘自行車，公共自行車已成為現今城市規劃中節能減碳的政策重點，為了讓公共自行車系統的資源效益最大化以及永續運輸之目的，對於影響公共自行車使用量之重要因素有深入探討的必要。 基於上述原因，為了找出影響公共自行車系統租借量之因素，本研究將以臺北市YouBike公共自行車系統159個租借站為研究樣本，分別對於租借量空間分佈與時間分佈進行分析，空間分析部分為構建全域型和區域型兩模式，其中全域型模式推估的參數，於各地區屬於均質的特性，採用模式為線性迴歸分析，而區域型模式可推估因地而異的參數，採用地理加權迴歸模式進行分析。將YouBike 159個租借站之現有能取得資料，分別以社會經濟、公共運輸、私人運輸相關變數來進行推估。而時間分析部分根據假日、平日租借量分開來討論，並且以迴歸分析之顯著變數作為分群依據，將租借站距離捷運站之距離分為遠、中、近三群， 將使用量以函數資料的方式進行分析，討論函數資料的平均函數、一次微分、二次微分、共變異數、相關係數，並進行函數資料之主成份分析，找出公共自行車租借站使用量之重要趨勢，並提供增加公共自行車使用量之相關對策。 結果顯示因地理加權迴歸可估計因地而異之參數值，其判定係數相較於全域型模式，為解釋能力較高的模式。而函數資料分析的部分，近距離群組需注意為早上通勤時大安區內的租借站，中、遠距離群組需注意早上8點到9點信義區內各租借站的狀況是否有無法負荷的狀況。假日注意近距離群組晚上6點的羅斯福新生南路口、捷運公館站，中距離群組則是傍晚5點到7點的捷運國父紀念館站、世貿二館站。遠距離的的熱門租車站點為早上9點到10點的臺北市立圖書館站、臺灣科技大學站，以及傍晚6點到7點的臺灣師範大學站、臺大資訊大樓站，基於以上之分析結果，給予相關之建議。

With the rapid growth of urban in recent years，the demand of transportation increases significantly. CO2 emissions and green house effect become a major challenge. Therefore, environmental protection and sustainable development attract everyone's attention. Governments begin to encourage people to walk, ride a bike as a tool between the public transit. Public bicycle system has become a policy priority in urban planning about CO2 reduction modern, in order to let the public bicycle system to maximize the effectiveness of resources and the objective of sustainable transport. Finding the determinants for the amount of the public bicycles usage is necessary. For these reasons, in order to find the key factors of public bicycle system usage. This study will use Taipei YouBike public bicycle system’s 164 rental stations as samples, doing the space distribution and time distribution analysis, respectively. The spatial analysis section build a global model and a local model. The global regression model is to estimate one set of parameters associated with explanatory variables to explain differences in the usage amounts of all rental stations, while local regression model can estimate parameters differ depending upon the correlation of neighboring rental stations. The variables we choose can be divided into three categories: socio-economic variables, public transport variables and private transport variables. The time analysis discuss the weekdays, weekends public bicycle usage amount separately, and with the significant variables in the regression analysis before as a basis for clustering. According to the distance from the MRT station categorized into far, middle, near three groups. After be used as a function data, We performed the functional data analysis and discussion of average function, a differential function, quadratic differential function, covariance, correlation coefficient, and functional data of principal component analysis to identify important trends in the amount of public bicycle usage, and give the relevant countermeasures. Compared with the local model, the result shows geographically weighted regression model’s R Square is better, because it can estimate the parameter values vary due to the different areas. Its explain ability is higher. And in the function data analysis part,the closed range groups need attention rental stations in Da-an district when the morning peak, the middle and long range groups should pay attention to 8:00 to 9:00 each rental stations in Xin-yi district if there have a overloading conditions. On weekend, the closed range groups need attention Roosevelt Xinsheng South Road, Gongguan MRT station at 18:00. The middle range groups should take care of Sun Yat-sen Memorial Hall MRT Station, Taipei World Trade Center 2 station at 17:00 to 19:00. Popular rental stations in long range groups Taipei Municipal Library Station, National Taiwan University of Science and Technology station from 9:00 to 10:00, and 6:00 to 7:00 in the evening in Taiwan Normal University Station, National Taiwan University information station building, based on the above results, given the relevant recommendations and management strategies.