Agent Asnalysis (AA) using a vision-driven agent simulation EVA (Exosomatic Visual Architecture) on a fine grid compared to visibility graph analysis (VGA) is not well known in Japan. Turner made EVA, a basic model of "natural movement" proposed by Hillier.
The authors conducted an encounter survey on weekday and weekend in 2017 in Kanayama, Nagoya, Aichi, and performed a correlation analysis between the counts of pedestrians on each street and various indicators by VGA and AA. As a result, One of the VGA indicators had a maximum correlation coefficient (single correlation) with the number of pedestrians of 0.320, but the an AA indicator showed a maximum of 0.800, and a strong correlation was confirmed (N=178).
So, in this paper, after examining multiple regression model selection by introducing VGA indicators, we will try model selection by introducing the AA indicators instead of the VGA indicators.
Our analytical framework explores the three contributions of accessibility, facility volume (land use intensity) and urban form, which are candidate factors (groups). In the first analysis, as the candidate factors (groups) of urban form, the connectivity and the global integration value as a VGA indicator, we use two kinds of footprint counts (Random Generated) in the second analysis, and (Station Generated) in the third analysis, both as the AA indicators. Each of Agent Analysis indicators means one simulation run result with 2000 agents in 2000 steps, but the generation points are randomly set in the walkable space in Random Generated case, and are set in each of eight station entrances and exits in Station Generated case.
The findings from our analyses are summarized below.
First, looking at the selection results of the first models that introduced the VGA indicators, for both weekday and weekend models, the variables are adopted without omission from the three factors of accessibility, facility volumes, and urban form. This supports the effectiveness as a factor of the VGA indicators for an urban form indicators in Kanayama dictrict. In addition, as the features of pedestrians’ distribution on weekend for weekday, the ground floor commercial building coverage rate are extracted.
Next, the second models that introduced the footprints (AA, Random Generated) indicator in both the weekday and weekend have improved both of multiple correlation coefficient and AIC values compared to the first model. In addition, the feature structure in the first model is preserved in this model. Moreover, the intensity of the footprints (AA, Random Generated) indicator is greater than that of the VGA indicators. This indicates that this indicator is a powerful urban form indicators that can improve and replace the VGA indicators.
Finally, the third models incorporating the footprints (AA, Station Generated) indicator in both the weekday and weekend shows the improvement of both multiple correlation coefficient and AIC compared to the second model. In addition, we can find the models also preserve the features in Kanayama also. However, although the intensity of the footprints (AA, Station Generated) indicator introduced as an urban form indicators became greater than the footprints (Random Generated) indicator, the intensity of the accessibility indicator (distance from the station) decreased. From this, it became clear that the footprints (Station Generated) indicator is not only strong as an urban form indicator but also has the property as an accessibility indicator.
抄録全体を表示