CN113487465B - City overlapping structure characteristic detection method and system based on label propagation algorithm - Google Patents

Abstract

The invention relates to the field of urban planning, and provides a method and a system for detecting urban overlapping structure features based on a label propagation algorithm, wherein the method comprises the following steps: obtaining city grids and processed trajectory data; carrying out weighted matching on the urban grids and the processed trajectory data to obtain a four-layer directed weighting network; inputting a graph division model into a four-layer directed weighting network for unsupervised training to obtain a four-layer urban community structure, and extracting an overlapping structure in the four-layer urban community structure; and constructing a measurement index through the point-of-interest data, and identifying the overlapped structure through the measurement index to obtain the land use characteristics and the space interaction mode of the overlapped structure. The method introduces the method for dividing the images in the network science into city planning, has better benefit, and can divide the city structure in batch and automatically; and the spatial interaction information of the city hidden in the urban resident activity can be fully excavated, and the land utilization characteristics and the spatial interaction relationship of the city are excavated.

Description

Method and system for urban overlapping structure feature detection based on label propagation algorithm

technical field

The invention relates to the field of urban planning, in particular to a method and system for detecting overlapping structural features of cities based on a label propagation algorithm.

Background technique

In the past three decades, the abundant labor force, good infrastructure and cheap land brought by the urbanization process have laid the foundation for rapid economic development. But it is unavoidable that many problems have arisen in the process of urbanization in our country. Especially for some provincial capital cities or metropolises, the urban problem is particularly serious. "Urban disease" is mainly manifested in traffic congestion, housing shortage, insufficient water supply, energy shortage, environmental deterioration, etc., which have caused a burden to the city and even restricted the development of the city. The development structure of a city is closely related to the life and economy of urban residents. The determination of the urban spatial structure through scientific means combined with human activities, and the provision of operational, scientific and reasonable spatial structure analysis methods, has become an important direction of digital city research.

The urban structure is becoming more and more complex and diverse, with obvious layers and overlaps. Urban areas at different levels have a relatively obvious layer-overlapping relationship. Exploring this layer-overlapping relationship from human activities is helpful to gradually grasp the urban spatial structure from the local to the whole. The interaction between the overlapping structure of the city and other plots is far greater than the interaction between itself and itself. It can be understood as the hub area of urban space interaction, and this interaction can be calculated by human behavior. Some experts have done related researches on urban structure division methods before, these methods can be divided into statistical survey-based methods and model-based methods, but there are few studies on urban structure hierarchy and urban overlap. Among them, the method based on statistical survey is combined with survey statistics and expert judgment. That is, in the process of urban structure delineation, based on the statistical results of field surveys, select a number of people who have a certain understanding of the city and are highly representative and Judging by authoritative experts. This method usually has a lot of subjectivity and high cost of time, manpower and capital. The model-based method delimits urban areas through scientific data analysis and big data mining with the support of crowdsourced geographic big data, and provides an operational, scientific and reasonable spatial optimization model. Crowdsource geographic data has the advantages of large data volume, strong current situation, abundant sources, and low cost. Based on the bottom-up collection characteristics of crowdsourced geographic data, researchers can easily obtain city-wide, massive and rich, personal-based spatiotemporal information, so as to achieve precise geographic analysis and modeling, and provide better insights for studying urban structure. Serve.

The above content is only used to assist the understanding of the technical solutions of the present invention, and does not mean that the above content is the prior art.

SUMMARY OF THE INVENTION

The main purpose of the present invention is to solve the technical problems of high subjectivity and high cost of time, manpower and capital in the prior art, and to detect and identify the features of overlapping hierarchical structures in cities.

In order to achieve the above purpose, the present invention provides a method for detecting overlapping structural features of cities based on a label propagation algorithm, comprising the steps of:

S1: Acquire city map data, taxi trajectory data, and point-of-interest data in the study area, divide the city map data into city units, and obtain an urban grid; preprocess the taxi trajectory data, and obtain the processed trajectory data;

S2: Perform weighted matching on the urban grid and the processed trajectory data to obtain a four-layer directed weighted network;

S3: Perform unsupervised training on the input graph partition model of the four-layer directed weighted network, obtain a four-layer urban community structure after the training is completed, and extract the overlapping structure in the four-layer urban community structure;

S4: Constructing a measurement index based on the point of interest data, identifying the overlapping structure through the measurement index, and obtaining land use characteristics and a spatial interaction pattern of the overlapping structure.

Preferably, step S1 is specifically:

S11: Process the city map data through GIS software, perform spatial fishnet analysis on the city area of the city map data, and divide the city area into the city grid; the city grid includes a plurality of 500mx500m cities grid cell;

S12: Eliminate the point data and invalid point data that are not in the urban area in the taxi trajectory data, and obtain the eliminated trajectory data;

S13: Extract the pick-up and drop-off point data in each piece of the excluded trajectory data, and the processed trajectory data is a set of the pick-up point data.

Preferably, step S2 is specifically:

S21: Match each of the pick-up and drop-off point data in the processed trajectory data with each of the urban grid units, compare each of the urban grid units as a graph node, and compare the difference between the urban grid units The number of interactions is analogous to the weight of the edge;

S22: The directed weighted network is composed of a plurality of the urban grid units and the interaction relationship between each of the urban grid units, and the interaction relationship is related to the number of interactions between the urban grid units;

S23: Perform route layering on the processed trajectory data, the first layer with the trajectory distance less than 3km, the second layer with the trajectory distance less than 5km, the third layer with the trajectory distance less than 9km, and the fourth layer with all the trajectory distances layer, respectively constructing corresponding directed weighted networks for the first layer, the second layer, the third layer and the fourth layer to obtain the four-layer directed weighted network.

Preferably, step S3 is specifically:

S31: Initialize the memory of each node in the graph partitioning model with the id of the corresponding node, and each node obtains a corresponding unique label;

S32: select a node as the listener node;

S33: All adjacent nodes of the listener node send their own unique tags to the listener node, and the listener node selects the most popular tag among all the tags received;

S34: Repeat steps S32-S33 for a total of n times, traverse all nodes, and obtain the most popular labels of all nodes;

S35: Perform post-processing on all the labels of each of the nodes to obtain the four-layer urban community structure. The division result of the four-layer urban community structure can be evaluated through an overlapping modularity function. The overlapping modularity function is specifically:

Among them, m is the weight sum of the edges in the network, A is the weighted adjacency matrix of the network, if there is an edge between node v and node w, then A _vw is the weight of the vw edge, otherwise it is 0; k _v , k _w are the out-degree weight sum of node v and the in-degree weight sum of node w, respectively, O _v , O _w are the number of communities to which node v and node w belong;

S36: Extract the overlapping structure in the four-layer urban community structure.

Preferably, in step S4, the measurement indicators include: richness, Simpson index and entropy measurement indicators;

Land use and functional type can be identified by the abundance; land mixing can be identified by the Simpson index and the entropy measure;

The formula for the richness is specifically:

F _i,l represents the enrichment index of the l-th type of POI in the ith plot, n _l,i represents the number of the l-th type of land use in the ith plot, n _i is the i-th plot The number of all POIs in . N _l is the total number of POIs of type l, and N is the total number of POIs in the entire study area;

The Simpson index and the entropy measurement index are represented by the Hill index, and the formula is specifically:

In the formula, D represents the value of the Hill index, and p _u represents the proportion of the u-th type of POI; when q=1, it represents the entropy. Ordered; q=2 is the inverse of the Simpson index, which measures the probability that two POIs randomly selected from an urban area belong to the same class; thus, it takes into account both the abundance of POIs and the Considering the relative abundance of different types of POIs, the lower the value, the higher the degree of mixed use of land; the higher the value, the lower the degree of mixed use of land.

Preferably, in step S4;

The characteristics of land use are the type of land use and the degree of land mixing, and functional areas are structures that exhibit certain functional attributes in urban areas, such as residential areas and scenic areas. The richness can be used to measure the functional structure and function of overlapping structures. Mixing degree; the POI mixing degree of overlapping structures can be calculated by the Simpson index and entropy measurement index, and the POI mixing degree can reflect the vitality of the plot;

The spatial interaction pattern represents the interaction between the overlapping area and neighboring communities. The strength of interaction is represented by the number of interactions, an interaction network is constructed, and people’s travel patterns are analyzed through the trajectory flow from one functional area to another. , the travel patterns such as morning rush hour from residential area to work area, holidays from residential area to leisure area and other interactive modes with fixed time rules, analyze the overlapping area interaction hotspot area by interaction strength and further use the richness Identify the functional structure of the interaction area.

An urban overlapping structure feature detection system based on label propagation algorithm, including the following modules:

The data acquisition module is used to acquire city map data, taxi trajectory data and point-of-interest data in the research area, divide the city map data into city units, and obtain city grids; preprocess the taxi trajectory data , obtain the processed trajectory data;

A four-layer directed weighted network acquisition module is used to perform weighted matching between the urban grid and the processed trajectory data to obtain a four-layer directed weighted network;

The overlapping structure extraction module is used to perform unsupervised training on the input graph partition model of the four-layer directed weighted network, obtain a four-layer urban community structure after the training is completed, and extract the overlapping structure in the four-layer urban community structure;

The identification module is configured to construct a measurement index based on the interest point data, identify the overlapping structure through the measurement index, and obtain the land use characteristics and spatial interaction pattern of the overlapping structure.

The present invention has the following beneficial effects:

1. The present invention adopts the method of analogical reasoning, and introduces the method of graph division in network science into urban planning, which has good benefits, and can divide the urban structure in batches and automatically;

2. The present invention can fully excavate the spatial interaction information of the city hidden in the activities of urban residents, at the same time pay attention to the overlapping structure existing in the network, and excavate its land use characteristics and its spatial interaction relationship.

Description of drawings

1 is a flowchart of a method according to an embodiment of the present invention;

FIG. 2 is a system structure diagram of an embodiment of the present invention;

The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

Referring to FIG. 1, the present invention provides a method for detecting urban overlapping structure features based on a label propagation algorithm, which is an extension method based on a model method. On the basis of previous research on spatial division models, the idea of network science is applied to urban division. Among them, and detect the overlapping structure existing in the city; this method is based on the idea of label propagation in graph division, adopts the method of analogical reasoning, aiming at the layering and overlapping of cities in the process of urbanization, and explores this layer from human activities. The overlap helps to gradually grasp the regional changes and spatial distribution of the urban spatial structure from the local to the whole; the present invention effectively introduces the concept of overlapping nodes in the complex network into the urban division, and combines the taxi trajectory data, which provides a better solution for long-term urbanization. The short-distance urban community structure is tested for the community, and the spatial interaction information of the city hidden in the activities of urban residents is fully excavated. At the same time, it pays attention to the overlapping structure existing in the urban community structure, and excavates its land use type and its spatial interaction relationship;

Specifically include the following steps:

S1: Acquire city map data, taxi trajectory data, and point-of-interest data in the study area, divide the city map data into city units, and obtain an urban grid; preprocess the taxi trajectory data, and obtain the processed trajectory data;

S2: Perform weighted matching on the urban grid and the processed trajectory data to obtain a four-layer directed weighted network;

S3: Perform unsupervised training on the input graph partition model of the four-layer directed weighted network, obtain a four-layer urban community structure after the training is completed, and extract the overlapping structure in the four-layer urban community structure;

S4: Constructing a measurement index based on the point of interest data, identifying the overlapping structure through the measurement index, and obtaining land use characteristics and a spatial interaction pattern of the overlapping structure.

In this embodiment, step S1 is specifically:

S11: Process the city map data through GIS software, perform spatial fishnet analysis on the city area of the city map data, and divide the city area into the city grid; the city grid includes a plurality of 500mx500m cities grid unit; in this embodiment, a total of 4853 urban grid units are obtained;

S12: Eliminate the point data and invalid point data that are not in the urban area in the taxi trajectory data, and obtain the eliminated trajectory data;

S13 : Extract the pick-up and drop-off point data in each piece of the excluded trajectory data, and the processed trajectory data is the set of the pick-up point data; in this embodiment, a total of 793253 pieces of pick-up and drop-off point data are acquired.

In this embodiment, step S2 is specifically:

S21: Match each of the pick-up and drop-off point data in the processed trajectory data with each of the urban grid units, compare each of the urban grid units as a graph node, and compare the difference between the urban grid units The number of interactions is analogous to the weight of the edge;

S22: The directed weighted network is composed of a plurality of the urban grid units and the interaction relationship between the urban grid units, and the interaction relationship is related to the number of interactions between the urban grid units; wherein, each Urban grid cells interact with several other urban grid cells;

S23: Perform route layering on the processed trajectory data, the first layer with the trajectory distance less than 3km, the second layer with the trajectory distance less than 5km, the third layer with the trajectory distance less than 9km, and the fourth layer with all the trajectory distances layer, respectively constructing corresponding directed weighted networks for the first layer, the second layer, the third layer and the fourth layer to obtain the four-layer directed weighted network;

In the specific implementation, the threshold of the route layering is determined according to the proportion of trajectories of each layer. The trajectories less than 3km, less than 5km and less than 9km account for 29.3301%, 51.1679% and 76.0485% of the total trajectories respectively; they can also be changed according to demand. The network structure of the four-layer directed weighted network.

In this embodiment, step S3 is specifically:

S31: Initialize the memory of each node in the graph partitioning model with the id of the corresponding node, and each node obtains a corresponding unique label;

S32: select a node as the listener node;

S33: All adjacent nodes of the listener node send their own unique tags to the listener node, and the listener node selects the most popular tag among all the tags received;

S34: Repeat steps S32-S33 for a total of n times, traverse all nodes, and obtain the most popular labels of all nodes;

In the specific implementation, when the four-layer directed weighted network input graph partition model is used for unsupervised training, the SLPA model parameters need to be set. In the graph partition model, the number of iterations of the model is set to 100, and the minimum number of divided communities is set to 3 , in order to keep the results of each division consistent, the random seed seed is set to 5140727168289296997. At the same time, after comparative testing and random seed value, the modularity fluctuation result of the community division result is less than 0.1, and the normal modularity value is 0.3 to 0.7. In addition, the r parameter that controls the output of overlapping communities is also determined by comparison, r∈{0.01,0.05,0.1,0.15,0.2,0.25,0.3,0.35,0.4,0.45,0.5}, through the comparison experiment, it is determined that when r=0.1 It is the inflection point of the modularity curve, and the final training parameter selects r=0.1, and the parameter selection can be changed according to the needs;

S35: Perform post-processing on all the labels of each of the nodes to obtain the four-layer urban community structure. The division result of the four-layer urban community structure can be evaluated through an overlapping modularity function. The overlapping modularity function is specifically:

Among them, m is the weight sum of the edges in the network, A is the weighted adjacency matrix of the network, if there is an edge between node v and node w, then A _vw is the weight of the vw edge, otherwise it is 0; k _v , k _w are the out-degree weight sum of node v and the in-degree weight sum of node w, respectively, O _v , O _w are the number of communities to which node v and node w belong;

In the specific implementation, the results of the overlapping modularity function of the four-story urban community structure fluctuate around 0.65, 0.6, 0.55, and 0.3, respectively. Considering the reasons for overlapping nodes and the length of the journey, it can be explained that the more overlapping nodes are identified, the better the area. The more chaotic, it may belong to multiple regions, the result of community division is not good, and the longer the distance, the larger the scale of community division, and the interaction of short distance will affect the division result; for overlapping nodes, the first community has the highest probability of labeling community, and so on;

S36: Extract the overlapping structure in the four-layer urban community structure.

In this embodiment, in step S4, the POI data is obtained through the AutoNavi API, and the data is reclassified; the POI data of each category is obtained by using the crawler method, a total of 622206 pieces of data, including POI name, latitude and longitude, Large category, medium category, small category, address and other attributes; also refer to the urban land use category, reclassify POI into 16 categories;

The measurement indicators include: richness, Simpson index and entropy measurement indicators;

Land use and functional type can be identified by the abundance; land mixing can be identified by the Simpson index and the entropy measure;

The formula for the richness is specifically:

F _i,l represents the enrichment index of the l-th type of POI in the ith plot, n _l,i represents the number of the l-th type of land use in the ith plot, n _i is the i-th plot The number of all POIs in . N _l is the total number of POIs of type l, and N is the total number of POIs in the entire study area;

The Simpson index and the entropy measurement index are represented by the Hill index, and the formula is specifically:

In the formula, D represents the value of the Hill index, and p _u represents the proportion of the u-th type of POI; when q=1, it represents the entropy. Ordered; q=2 is the inverse of the Simpson index, which measures the probability that two POIs randomly selected from an urban area belong to the same class; thus, it takes into account both the abundance of POIs and the Considering the relative abundance of different types of POIs, the lower the value, the higher the degree of mixed use of land; the higher the value, the lower the degree of mixed use of land.

In this embodiment, in step S4;

The characteristics of land use are the type of land use and the degree of land mixing, and functional areas are structures that exhibit certain functional attributes in urban areas, such as residential areas and scenic areas. The richness can be used to measure the functional structure and function of overlapping structures. Mixing degree; the POI mixing degree of overlapping structures can be calculated by the Simpson index and entropy measurement index, and the POI mixing degree can reflect the vitality of the plot;

The spatial interaction pattern represents the interaction between the overlapping area and neighboring communities. The strength of interaction is represented by the number of interactions, an interaction network is constructed, and people’s travel patterns are analyzed through the trajectory flow from one functional area to another. , the travel patterns such as morning rush hour from residential area to work area, holidays from residential area to leisure area and other interactive modes with fixed time rules, analyze the overlapping area interaction hotspot area by interaction strength and further use the richness Identify the functional structure of the interaction area.

Referring to Fig. 2, the present invention provides an urban overlapping structure feature detection system based on a label propagation algorithm, including the following modules:

The data acquisition module 10 is used for acquiring city map data, taxi trajectory data and point-of-interest data in the research area, dividing the city map data into city units, and obtaining an urban grid; pre-predicting the taxi trajectory data processing to obtain the processed trajectory data;

A four-layer directed weighted network acquisition module 20 is used to perform weighted matching between the urban grid and the processed trajectory data to obtain a four-layer directed weighted network;

The overlapping structure extraction module 30 is used to perform unsupervised training on the four-layer directed weighted network input graph division model, obtain a four-layer urban community structure after the training is completed, and extract the overlapping structure in the four-layer urban community structure;

The identification module 40 is configured to construct a measurement index based on the point of interest data, identify the overlapping structure through the measurement index, and obtain land use characteristics and spatial interaction patterns of the overlapping structure.

It should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, It also includes other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments. In a unit claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. do not denote any order, and these words may be construed as identifications.

The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields , are similarly included in the scope of patent protection of the present invention.

Claims (4)

1. A city overlapping structure feature detection method based on a label propagation algorithm is characterized by comprising the following steps:

s1: obtaining city map data, taxi track data and interest point data in a research area, and carrying out city unit division on the city map data to obtain city grids; preprocessing the taxi track data to obtain processed track data;

s2: carrying out weighted matching on the urban grids and the processed trajectory data to obtain a four-layer directed weighting network;

step S2 specifically includes:

s21: matching each boarding and alighting point data in the processed track data with each city grid unit, simulating each city grid unit into a graph node, and simulating the number of interactions among the city grid units into the weight of an edge;

s22: the directed weighting network is composed of a plurality of city grid units and interactive relations among the city grid units, and the interactive relations are related to the interactive times among the city grid units;

s23: path layering is carried out on the processed track data, a first layer is used when the track path is less than 3km, a second layer is used when the track path is less than 5km, a third layer is used when the track path is less than 9km, all track paths are fourth layers, corresponding directed weighting networks are respectively constructed on the first layer, the second layer, the third layer and the fourth layer, and the four layers of directed weighting networks are obtained;

s3: performing unsupervised training on the four-layer directed weighting network input graph division model, obtaining four-layer urban community structures after the training is finished, and extracting an overlapping structure in the four-layer urban community structures;

step S3 specifically includes:

s31: initializing the memory of each node in the graph partitioning model by using the id of the corresponding node, and obtaining a corresponding unique label by each node;

s32: selecting a certain node as a listener node;

s33: all adjacent nodes of the listener node send own unique labels to the listener node, and the listener node selects the most popular label from all the received labels;

s34: repeating the steps S32-S33 for n times, traversing all the nodes and obtaining the most popular labels of all the nodes;

s35: post-processing all the labels of the nodes to obtain the four-layer urban community structure, and evaluating the division result of the four-layer urban community structure through an overlapping modularity function, wherein the overlapping modularity function specifically comprises the following steps:

wherein m is the sum of the weights of the edges in the network, A is a weighted adjacency matrix of the network, and if an edge exists between a node v and a node w, A is _vw The weight of vw edge is, otherwise, 0 is adopted; k is a radical of _v ，k _w Respectively, the out-degree weight of the node v and the in-degree weight of the node w, O _v ，O _w Respectively to which node v and node w belongThe number of communities of;

s36: extracting an overlapping structure in the four-layer city community structure;

s4: constructing a measurement index through the point of interest data, and identifying the overlapped structure through the measurement index to obtain the land use characteristics and the space interaction mode of the overlapped structure;

the land use characteristics are land use types and land mixing degrees, and the space interaction mode represents the interaction condition of the overlapping structure and the adjacent community;

in step S4, the measurement index includes: richness, simpson index and entropy measure index;

the land use condition and the function type can be identified through the richness; the land mixing condition can be identified through the Simpson index and the entropy measurement index;

the formula of the richness is specifically as follows:

F _i,l denotes the enrichment index, n, of POIs of class I in the ith plot _l,i Representing the number of class I land use types in the ith plot, n _i Is the number of all POIs in the ith parcel; n is a radical of _l N is the total number of POIs of class i, and N is the total number of POIs in the whole research area;

the simpson index and the entropy measurement index are expressed by a hill index, and the formula specifically comprises:

in the formula, D represents the value of the Hill index, p _u Representing the proportion of u-type POI; when q is 1, the representative entropy is shown, the higher the value is, the more disordered the POI type distribution is, and the lower the value is, the more orderly the POI type distribution is; q is the inverse of the happson index when q is 2, measured from oneProbability that two randomly selected POIs in a city area belong to the same category; therefore, the abundance of the POI and the relative abundance of different types of POI are considered, and the lower the value is, the higher the mixed utilization degree of the land is, and the higher the value is, the lower the mixed utilization degree of the land is.

2. The method for detecting urban overlapping structure features based on the label propagation algorithm according to claim 1, wherein the step S1 specifically comprises:

s11: processing the urban map data through GIS software, performing spatial fishing net analysis on urban areas of the urban map data, and dividing the urban areas into urban grids; the city grid comprises a plurality of 500mx500m city grid cells;

s12: removing the point data and invalid point data which are not in the urban area in the taxi track data to obtain removed track data;

s13: and extracting the boarding and alighting point data in each eliminated track data, wherein the processed track data is a set of the boarding and alighting point data.

3. The city overlap structure feature detection method based on label propagation algorithm according to claim 1, characterized in that in step S4;

the functional area is a structure showing certain functional attributes in urban areas, and comprises the following steps: the functional structure and the function mixing degree of the overlapping structure can be measured through the richness; the POI mixing degree of the overlapped structure can be calculated through the Simpson index and the entropy measurement index, and the POI mixing degree can reflect the vitality of the land;

expressing the strength of interaction through the number of times of interaction, constructing an interaction network, and analyzing the travel mode of people through the trajectory flow from one functional area to another functional area, wherein the travel mode comprises the following steps: and the interactive mode with a fixed time law is that the early peak is from the residential area to the working area, and the holiday is from the residential area to the leisure area, the interactive hotspot area of the overlapped structure is analyzed through the interactive intensity, and the functional structure of the interactive area is further identified by using the richness.

4. A city overlapping structure feature detection system based on a label propagation algorithm is characterized by comprising the following modules:

the data acquisition module is used for acquiring city map data, taxi track data and interest point data in a research area, and performing city unit division on the city map data to obtain city grids; preprocessing the taxi track data to obtain processed track data;

the four-layer directional weighting network acquisition module is used for carrying out weighted matching on the urban grid and the processed track data to obtain a four-layer directional weighting network;

the steps for obtaining the four layers of directional weighting networks are as follows:

s21: matching each boarding and alighting point data in the processed track data with each city grid unit, simulating each city grid unit into a graph node, and simulating the number of interactions among the city grid units into the weight of edges;

s22: the directed weighting network is composed of a plurality of city grid units and interactive relations among the city grid units, and the interactive relations are related to the interactive times among the city grid units;

s23: path layering is carried out on the processed track data, a first layer is used when the track path is less than 3km, a second layer is used when the track path is less than 5km, a third layer is used when the track path is less than 9km, all the track paths are fourth layers, corresponding directed weighting networks are respectively constructed on the first layer, the second layer, the third layer and the fourth layer, and the four layers of directed weighting networks are obtained;

the overlapping structure extraction module is used for carrying out unsupervised training on the four-layer directed weighting network input graph division model, obtaining four-layer urban community structures after the training is finished, and extracting an overlapping structure in the four-layer urban community structures;

the step of extracting the overlapping structure in the four-layer city community structure comprises the following steps:

s31: initializing the memory of each node in the graph partitioning model by using the id of the corresponding node, and obtaining a corresponding unique label by each node;

s32: selecting a certain node as a listener node;

s33: all adjacent nodes of the listener node send own unique labels to the listener node, and the listener node selects the most popular label from all the received labels;

s34: repeating the steps S32-S33 for n times, traversing all the nodes, and obtaining the most popular labels of all the nodes;

s35: post-processing all the labels of the nodes to obtain the four-layer urban community structure, and evaluating the division result of the four-layer urban community structure through an overlapping modularity function, wherein the overlapping modularity function specifically comprises the following steps:

wherein m is the weight sum of edges in the network, A is the weighted adjacent matrix of the network, and if an edge exists between the node v and the node w, A is _vw The weight of vw edge is, otherwise, 0 is adopted; k is a radical of _v ，k _w Respectively, the out-degree weight of the node v and the in-degree weight of the node w, O _v ，O _w The number of communities to which the node v and the node w belong respectively;

s36: extracting an overlapping structure in the four-layer city community structure;

the identification module is used for constructing a measurement index through the interest point data, identifying the overlapped structure through the measurement index and obtaining the land use characteristics and the space interaction mode of the overlapped structure;

the land use characteristics are land use types and land mixing degrees, and the space interaction mode represents the interaction condition of the overlapping structure and the adjacent community;

the measurement index includes: measuring indexes of richness, Simpson indexes and entropies;

the land use condition and the function type can be identified through the richness; the land mixing condition can be identified through the Simpson index and the entropy measurement index;

the formula of the richness is specifically as follows:

F _i,l denotes the enrichment index, n, of POIs of class I in the ith plot _l,i Representing the number of type I land use types in the ith plot, n _i Is the number of all POIs in the ith parcel; n is a radical of hydrogen _l N is the total number of POIs of class i, and N is the total number of POIs in the whole research area;

the Simpson index and the entropy measurement index are expressed by a Hill index, and the formula is as follows:

in the formula, D represents the value of the Hill index, p _u Representing the proportion of u-type POI; when q is 1, it represents entropy, and higher values indicate that the distribution of POI species is more disordered, and lower values indicate that the distribution of POI species is more ordered; q is 2, which is the inverse of the fortunate pson index, which measures the probability that two POIs randomly selected from a metropolitan area belong to the same category; therefore, the abundance of the POI and the relative abundance of different types of POI are considered, and the lower the value is, the higher the mixed utilization degree of the land is, and the higher the value is, the lower the mixed utilization degree of the land is.

Images