Forecasting the Degree of Crowding in Urban Public Open Space upon Multi-source Data

As the urban population has risen steadily, it plays a major role in urban security and traffic control to make an accurate prediction of the degree of crowing in urban public open space. In this paper we propose an approach to predict the degree of crowding in urban public open space based on the multi-source data. Here we divide an urban public open space into a number of areas while a day into four time periods. The grid-sampling method is given to obtain the pedestrian volume index of different historical time slots in an area according to the number of Wechat's users. The spatio-temporal features of an area are determined by urban multi-source data including the maps, POI data, public comments data, area property data, meteorological data etc. Based on the historical pedestrian volume index and spatio-temporal features, we use the Takagi-Sugeno (T-S) fuzzy neural network model to forecast the pedestrian volume index of an area, and then converts the pedestrian volume index into the degree of crowding by the given EDCC algorithm (the evaluation of degree of crowding by clustering). This paper uses the middle portion of Shanghai Binjiang Avenue as a study case to predict the place's degree of crowding within a week and the result verifies the effectiveness of the proposed approach.