Generating Shop-Around Agent Model Using Aggregated Partial Information

Abstract

This study discusses a statistical modeling approach to simulate shopping around behavior agents using two types of macro-aggregate partial information: the number of visitors by zone and the number of passers-by by street link. Parameter estimation is described for two district cases, Osu and Nishi-Shinjuku, and simulation performance is examined. In the Nishi-Shinjuku case, we also present the results of policy effect estimation. This agent model introduces the shortest path principle into the Markovian shopping around behavior model and uses the inverse transformation of the binomial logit to perform probabilistic selection of routes. This allows for the possibility of extending the model to handle not only objective distances but also subjective distances. Future work is needed to improve the deviation rate, which indicates the degree of fitting.