小汽車駕駛人路線選擇行為模式之研究

Abstract

本研究在觀察影響小汽車駕駛人從事路線選擇行為時所考慮之重要變 數，並以台北都會 區內小汽車駕駛人為研究對象，配合所設計之調查方 法與問卷，蒐集其路線選擇行為過程中相關資料與數據，以構建模式並分 析之；同時根據小汽車駕駛人之旅次起迄點加以分類，建立適用於台北市 都會區小汽車駕駛人之路線選擇行為模式。 本研究採用二階段問卷調 查，第一階段針對台北都會區內小汽車駕駛人抽樣調查其選擇路線時所考 慮之重要變數，並以第一階段所得結果設計第二階段問卷。第二階段問卷 設計以蒐集顯示性偏好數據與敘述性偏好數據為主要設計方向。顯示偏好 數據包括駕駛人實際使用路線與可能替代路線資料；敘述偏好數據則是駕 駛人在面對特定運輸情境時，選擇其所偏好之路線，同時調查受訪者年齡 、性別、所得等個人基本資料以及使用車輛情況、旅次時間等旅次特性資 料。 本研究構建模式時，係採用敘述性偏好所設計之路線選擇問題 ，並配合個人基本資料及旅次特性資料，應用羅吉特模式(Logit model) 理論建立路線選擇行為模式。最後為一實證研究，係利用專家系統分析內 知識擷取法之應用工具 ID3演算法，建立一台北都會區內小汽車駕駛人路 線選擇之應用實例。 本研究結果可獲得以下幾點結論： 1.由資料整 理結果顯示，經由問卷調查所得之十七項個人基本資料及旅次特性資料中 ，經過交叉分析並檢定其相關性而知，在「職業」項目中具有最多相關資 料，計有八項，在「是否保全險」及「性別」二項相關資料最少，僅有一 項，分別為「車齡」及「職業」。2.各分區模式構建之結果顯示，小汽車 駕駛人在路線選擇時重視之因素，以「可行駛速度」及「實際行駛距離」 兩項最重要，其次為「交通資訊提供」、「左轉次數」等項目。在個人基 本資料部份以「性別」、「年齡」、「交通費用支出」及「每月所得」較 有影響，旅次特性資料以「旅次時間」、「車齡」、「是否保全險」、「 排氣量」、「車輛所有人」及「超過最大容忍延誤時，所採取方式」等變 數較有影響。在各分區模式之ρ2值大多表現良好，顯示所測定的模式對 數據有相當高的配合能力。 3.實證研究之結果顯示，於各分區所得歸納 決策樹來看，控制屬性個數在三至四個內，以減低決策樹之複雜前提下， 各區之預測成功率皆在五成以上，顯示應用專家系統分析之知識擷取法， 對預測於駕駛人路線選擇行為具有實用性。 The purpose of this study is to observe the important variables that will effect the route choice behavior of car drivers in Taipei Urban Area. We collected the related data of route choice behavior and then classify them by trip origin and destination, to construct proper route choice behavior models of car drivers in Urban Area Taipei. In this study we used two-step survey. First we collected the important variables that will effect the route choice behavior of car drivers in Taipei Urban Area. Then we used previously results to design next survey. In this survey, we collected the revealed preference data and stated preference data. The revealed preference data includes real route and substitute route; The stated preference data means the route which car drivers select when face an especially transportation situation. We also collected some personal basic data of answers just as age, sex, income and so on; some data of using car situation and trip characters. This study used Logit model to construct route-choice behavior models according to different route choice set and personal basic data and trip characters data. Finally, we construct a positive study. It used the ID3 algorithm that is one kind of knowledge acquisition of expert analysis system to construct a really case study in Taipei Urban Area. According to this study, we can find the following results: 1. After cross analysis and testing of statistical hypothesis the personal basic data and trip characters data totally 17 items, we find that 'occupation' is most related with other items, and that 'fully insurance' and 'sex' are least related with other items. 2. Generally speaking, running speed and really running distance are most important variables in deciding car drivers' route choice behavior, then are traffic information provided and left-turn numbers. In the personal basic data, sex, age, traffic expense and income are more important variables. In the trip characters data, trip time, car age, fully insurance, displacement (c.c.), owner of car, and changes route situations are more important variables. The Rho-squre value (ρ2) appears good in every model,they show that these models are proper in Taipei Urban Area. 3. In the positive case study, if we control the attributes to 3 or 4, the forecast successful rates of inductive decision tree are more than 50%, they show that ID3 algorithm is a utility method in forecasting route-choice behavior of car drivers.