Street choice logit model for strolling visitors in shopping district

Streets should be designed as spaces in town where people feel both efficiency for travelling and attractiveness for rambling. By predicting people's distribution and walking around activity in cities, urban planning will become more successful. We investigate a shopping district arround the suburb railway station in Japan, and propose two models for predicting people's activity. The first model is a pedestrians distribution prediction (or postdiction) model by multiple regression analysis using space syntax indices of urban fabric and people's distribution data obtained from the field-survey. The second model is a street choice model uging Logit model. We performed a questionnaire survey on the field and investigated 46 visitors' strolling routes and obtained totally 1211 street choices in thier routes. A utility function, sum of weighted space syntax indices and other indices, is made and the parameters for weights were estimated by maximum likelihood. These models take account of both street networks (Integration Value, distance from the destination) and other spactial compositions (amount of pedestrians, cars and shops, road width, altitude). First model explained the characteristics of the street where a lot of people tend to walk or stay. The parameters of large weight were "near from the station", "wide", "with many stores", and "near to centricity (high Int.V)". Second model clarified differences of weights of street choice parameters between the attributes such as gender, the existence of destinations, number of people, and so on. There are many findings such as followings; males tend to choose the street which has relatively low Int.V, many cars; couples tend to choose the streets which have few pedestrians; people without destination tend to choose the streets which have high Int.V, a lot of pedestrians, narrow width, and few shops; people who move in a group tend to choose the streets which have few cars, many shops, wide width.