公路車輛起迄抽樣調查與不完整資料處理之研究

Abstract

運輸規劃中，交通起迄流量是基本之資料。交通起迄流量調查方法有很多 ，包括家戶訪問、路邊訪問調查法等等，於開放性路網（即有穿越性旅次 ）中，車輛牌照登錄法為較可行的方法。然而在交通流量非常大時進行母 體調查並不容易，因此抽樣調查將是一個可試行之辦法。但在起迄調查研 究中，若抽樣比例微小，進而將此抽樣資料比對，所得到的可用資料量相 當的小，會因此產生不完整資料(incomplete data) 的問題。因此，本文 將討論如何補足缺失值(missing data)之問題，並構建一不須任何型態分 佈假設之處理不完整資料的演算模式，進一步利用此資訊進行抽樣分析， 探討由抽樣得到的樣本，其代表母體之程度為何。研究內容包含不完整資 訊處理､抽樣分析及模式優劣評估指標的建立。 從本文之測試結果顯示 ，抽取之樣本與母體型態類似。故在實際調查時，只要以適當比例抽樣調 查，再經過不完整資訊之處理、等比例放大，即可得到與母體型態相同的 起迄表。如此便可節省相當的人力､經費，並可作為以後調查者之參考。 In transportation planning, the origin-destination trip flow is the most basic data. There are many kinds of methods to investigate origin-destination trip flow, including home interview surveys and road-side interview surveys and so on. For an open network, one of the most useful method is vehicle- license-mount method. Howevre, it is not easy to progress investigating population when the traffic is heavy. This is why sampling is useful. There is only a small portion of useful information obtained in O-D survey after matching the small sampling data. So, it often turns out incomplete data. This research will build a model that does not need any distribution, discuss how to supplement the incomplete data, analyze the imputed data set and explore the extension of the samples to represent the population. The content of this paper includes treatment of incomplete data, sampling analysis and setting up an index to test the model. The results show that the pattern of sampling data is similar to that of populations. Therefore, we can gather the data by system sampling, match the data to get O- D information and impute the incomplete data if necessary, amplify the complete table to estimate the population O-D table. This process is easy to implement and cost can be lowered.