碩士 / 中原大學 / 土木工程學系 / 88 / On the changing of economic growing, urban expanding and household life transforming in urban area, the travel behavior of individuals is more complex and unpredicted than before. In the past, researchers analyzed travel behavior according to single trips. Not only did it ignore individuals’ economic consideration, nor it not reflects the rationality during travel decision-making. In this study, the activity-based theory is incorporated into analyses, and trip chains are included in addition to the conventional trip units to explore the travel characteristic. The travel data in use are from year 1992 and 1997 travel surveys of Taipei metropolitan area; the characteristics analysis was conducted and the urban travel behavior forecasting models were developed, in hope to understand the influence of personal/household factors and the household-member interactionship to the travel behavior.
For establishing travel behavior forecasting models, the previous studies usually developed models based on conventional statistic methods. However, Statistic models require a few assumptions and are normally constrained by the linearity, which is not suitable of explaining the complexity of travel behavior. Recent efforts were devoted to applying neural network on transportation studies, and sound results were reported. Both neural network and logistic regression methods are used in this research; this study also tried to explore the influencing factors and compared the analysis results as well as forecasting effectiveness between models.
From the analysis results, individuals tend to chain trips of various purposes during activity participation, in order to accomplish duties under temporal and spatial constraints. This chaining phenomenon is growing with time changes. The usually picked personal/household socio-economic attributes are observed shaping trip chaining; certain factors such as the presence of young children show even stronger influence on household members’ chaining behavior. The time-series study was conducted and the above influence was increasing significantly.
By comparing two models, the analysis results show that different models are good for different problem patterns. For direct and complex chaining work journeys, neural network approaches give better forecasting results; which may state that the hidden layer and the neuron of neural network are suitable of simulating complex relationship. On the contrary, two models display little variation on forecasting abilities for non-work chaining journeys; which may imply a less complicated relationship between variables in such travel types, or the functional forms with less significant nonlinearly.
Identifer | oai:union.ndltd.org:TW/088CYCU0015001 |
Date | January 2000 |
Creators | Lin, Chun-Hsien, 林俊賢 |
Contributors | Liao, Yu-Chun, 廖祐君 |
Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
Language | zh-TW |
Detected Language | English |
Type | 學位論文 ; thesis |
Format | 111 |
Page generated in 0.0076 seconds