A Modeling Framework For Urban Mobility Based On Data Fusion

For a long time,the research of urban mobility has been the foundation of solving problems of public health,public safety and urban planning.Constructing the individual-based urban mobility model to carry out the related research is helpful to study the internal mechanism of the related problems and further carry out the scenario analysis.According to the data source utilized in recent urban mobility model,firstly,traditional travel survey data has limitations such as low update frequency,low sampling rate and high expense.Secondly,the large-scale trajectory data shows data bias and lacking of social demographic attribute information.Moreover,the census data which covers the entire population is short of human mobility information.As a consequence,the existing individual-level urban mobility model based on a single travel data sources will also be affected by its data source and will have corresponding inherent defects.Aiming at the problems above,our research bring up a modeling framework for individual-based urban mobility in order to integrate the advantages of different data sources and make up for its limitations.This modeling framework uses the approach of generating synthetic population,performing the home/work location identification algorithm to different travel data sources,introducing multiple travel data sources to the synthetic population by adopting the home location as latent consensus,and generating individual mobility based on travel data sources.As a result,it can generate an urban mobility model covering the entire population and possessing social demographic attribute information.On the basis of this,a trip generation algorithm based on spatio-temporal choices is proposed in our research.This algorithm generates daily activity chains randomly for the synthetic individuals by choosing activity for both fixed time slots and flexible time slots.Meanwhile,basing on hotspot detection and topic model,it will select an area of “other” activity for individuals and eventually construct the daily trip chains for every individual through exploration and preferential return model.Our research take the mobile phone location data from Shenzhen,Shanghai,the travel survey data from Shenzhen and so on as the research data source to verify the effectiveness of the method proposed in this study.The experimental results indicate that:(1)Compared to the existing model,this study can also generate reasonable trip chain and social demographic attribute for individuals outside the coverage of large-scale trajectory data,which can help to improve the defect of the individual-level urban human mobility model based on single travel data source;(2)Individuals under the same working group tend to have similar overall activity patterns,thus contributing to the better use of the large-scale trajectory data,which covers only a part of the population,to construct activity chains for the whole population.(3)Most of the individuals’ “other” activities are distributed within the hotspots of the “other” activities nearest to their home/work location,moreover,the use of topic models to explore the co-occurrence of “other” activity hotspots will help to improve the effect of selecting locations of “other” activities for the individual.The main contribution of this study mostly lies in:(1)An urban mobility modeling framework based on data fusion is proposed to improve the defects of the existing urban mobility model based on a single travel data source,and to provide a new idea for the combination of traditional data and big data;(2)Bring up an algorithm based on spatio-temporal choices which can effectively capture the fixed activity pattern and flexible activity pattern of different working groups and is helpful to use limited samples to construct the activity chain for the whole population;(3)Analyze the spatial distribution characteristics of individuals’ “other” activities,and introduce the topic model to further explore its distribution patterns,which is helpful to improve the effect of reconstruction of individuals’ “other” activities.To sum up,the result of this research will provide new ideas of the establishment of urban mobility model and help to improve the modeling effect,thus contributing to the reference of a more scientific and reliable urban human mobility model in the field of public health,public safety and urban planning.