Abstract
With the use of individual-level travel survey datasets describing the detailed activities of households, it is possible to analyze human movements with a high degree of precision. However, travel survey data are not without quality issues. Potential exists for origins and destinations of reported trips not to be geo-referenced, perhaps due to misreported information or inconsistencies in spatial address databases, which can limit the usefulness of the survey data. From an analytical standpoint, this is a serious problem because a single unreferenced stop in a trip record in effect renders that individual’s data useless, especially in cases where analyzing chains of activity locations is of interest. This paper presents a framework and basic computational approach for exploring unlocatable activity locations inherent to travel surveys. Derived from recent work in developing a network-based, probabilistic time geography, the proposed methods are able to estimate the likely locations of missing trip origins and destinations. The methods generate probabilistic potential path trees which are used to visualize and quantify potential locations for the unreferenced destinations. The methods are demonstrated with simulated survey data from a smaller metropolitan area.
Original language | American English |
---|---|
Journal | Computers, Environment and Urban Systems |
Volume | 36 |
DOIs | |
State | Published - Nov 1 2012 |
Keywords
- time geography
- travel surveys
- spatial behavior
- geocoding
- visualization
- probability
- error
- networks