Cascade Urban Area Analyze Framework For Clustering Urban Areas Using Ride-hailing Data

Urban areas play a major role in human life.Reasonable clustering urban areas can help people understand a city better.However,it is different to cluster urban areas for the relationship between areas is complex and changeful.With continued increases in innovation transportation services,big data and its related technologies have been widely applied to analyze urban areas.Specially,ride-hailing data is considered to cluster and analyze urban areas' relationship better due to it records a series of GPS points and other information such as order information,time.Thus,it has gradually attracted attention in region function,traffic planning,etc.Using ride-hailing data,complex network theory has been used to analyze urban areas,where grids or small zones are nodes and the connections between nodes are edges.A constructed network provides a new insight to analyze area's relationship.Understanding of spatial relationship between regions can help people correctly understand the implicit relationship between areas.At the same time,it also put forward suggestions for urban planning.In general,people usually use division of administrative regions to understand area relationship.However division of administrative regions does not immediately reflect the actual size and impact of its economic and transport.The existing research that is used to analyze the spatial relationship between urban areas considers an aspect,such as commuting frequency.However,there are some works to comprehensively analyze the relationship between urban areas.To fill this gap,this paper first proposes a comprehensive index and on this basis proposes a cascade framework to cluster urban areas.The main contents as follows:(1)First,this paper first proposes a new index named ABI(area behavior index)to analyze the spatial relationship between urban areas.When ABI studies the spatial relationship between areas,it does not consider the commuting frequency and the time cost of commuting between areas,and also considers the comparison of output matrix corresponding to every area.(2)Secondly,based on ABI,this paper develops a comprehensive,multi-pronged framework named CUAAF(cascaded urban area analyze framework)based on ride-hailing data for urban areas clustering.The framework consists of three indexes,in addition to the above ABI,including AAPI(area active pattern index)and AFI(area function index).With this approach,this paper has synthesized area spatial relationship,area active patterns and area functions.Specifically,the city is first segmented into multiple geographic grids.Secondly,the correlation between any two grids is evaluated using one of the indexes.Then,a grid network is constructed where grids are taken as nodes and the correlation between grids are taken as the weight of connecting nodes.Finally,the grid network is analyzed by the Louvain algorithm and the clustering areas are obtained.After a clustered area is got using an index,the area can be inputted into the CUAAF again to analyze further detailed information using any one of the three indexes.The process and results are as follows:To fully understand urban areas' spatial relationship,we use ABI to analyze Shijiazhuang.The results show that there are some different administrative regions belong to a same community,which indicates they have tight spatial associations.Thus,it is not accurate enough to understand area relationship only depend on administrative district division.This paper use AAPI respectively to analyze the Shijiazhuang and its core area.Results show that there are multiple time-based human flow patterns in the city.Moreover,urban areas with the same time-based human flow patterns are geographically dispersed.With the development of urban economy,area functions are not single.Area function mixing is common for most urban areas.To understand area functions accurately,this paper use AFI to analyze the core of Shijiazhuang.The results show that there are different area functions corresponding to POI proportions.Furthermore,this paper take one of the above results input into the CUAAF again to obtain further detailed areas using any one of the three indexes.This step is named cascading index.A different cascading sequence reflects different concerns of user.In this paper,this paper cascade AAPI and AFI to analyze Shijiazhuang core area.Thus,there are two sequences.One is to use AAPI first then AFI for analysis.The other case is the opposite.The former results show that there are different area functions in a same area active pattern and the latter results show that there are also different area active patterns in a same area function.(3)Lastly,this paper also applies the three indexes on weekdays and weekends respectively.The spatial relationship between areas is changed on weekdays and weekends for people's different lifestyle.In addition,the maximum areas with same area active pattern in weekends larger than weekdays for most people prefer to go around at weekends.Finally,the area major functions are variable in the two periods.These works justify the robustness and sensitivity of our approach.