A gosper-island-based hierarchical spatial community mining method

Abstract

The application discloses a hierarchical spatial community mining method based on Gosper-island and belongs to the field of spatial data mining. First, the application carries out hierarchical partitioning of space by adopting Gosper-island, uses hierarchical partitioning coding to record the spatial hierarchical constraint of a partitioning unit, and uses Gosper space filling coding to record the spatial proximity constraint of the same hierarchical partitioning unit. Then, the hierarchical spatial network is constructed by taking the partitioning unit as a network node, the network is organized in the form of a network adjacency matrix, the network adjacency matrix heat map is drawn according to the Gosper coding sequence of the node, and finally, the hierarchical matrix block structure of the network heat map at different levels is extracted from top to bottom to obtain the hierarchical spatial community structure meeting the spatial hierarchical constraint. The application conforms to the hierarchical characteristics of the hierarchical spatial community and can effectively avoid the problems caused by the partitioning hierarchical tree.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of spatial data mining, in particular to a hierarchical spatial community mining method based on Gosper-island (a hierarchical hexagonal grid with a hole diameter of 7). BACKGROUND

[0002] The relationship between nodes and communities in the hierarchical community structure is a one-to-many relationship, and nodes belong to different levels of communities. There is also a belonging relationship between different levels of communities. Hierarchical spatial community can better reflect the multi-level (scale) characteristics of the real world, and its detection results can better assist people in observing and studying geographical objects from a multi-level (scale) perspective. A large number of facts show that this is a more efficient cognitive method that conforms to the cognitive law. Hierarchical spatial community is an important means of comprehensive cognition of spatial network structure. However, the traditional hierarchical community detection method often ignores the influence of spatial constraints when detecting spatial networks. Spatial constraints and network constraints jointly affect the hierarchical functional zoning of spatial networks. At present, on the one hand, the field of complex networks has not conducted much research on hierarchical community structure detection. Most of the research focuses on the detection of time hierarchical community structure of dynamic complex networks, and the selection of hierarchical levels (or resolution) is often difficult to determine. Existing methods cannot meet the requirements of hierarchical division of spatial networks considering spatial proximity. On the other hand, the research on spatial network community structure detection considering spatial constraints is very scarce. Community structure detection cannot take into account the hierarchy and proximity in space, and most research equates hierarchical spatial community to clustering tree of divisive community detection. Therefore, it is necessary to analyze the spatial proximity constraint in combination with the characteristics of hierarchical spatial community detection of spatial networks, and design a new method suitable for hierarchical spatial community detection of spatial networks. SUMMARY

[0003] In order to mine the hierarchical spatial community structure characteristics of spatial networks, the present application proposes a hierarchical spatial community mining method based on Gosper-island. This method is simple and easy to implement, and conforms to the hierarchical characteristics of hierarchical spatial community, which can effectively avoid the problems caused by the division of hierarchical tree.

[0004] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0005] A hierarchical spatial community mining method based on Gosper-island, comprising the following steps:

[0006] (1) The spatial region is divided by using a regular hexagon, the hexagonal grid units after division are hierarchically organized by using Gosper-island, and the hierarchical nesting relationship between the hierarchical hexagonal grids in the spatial region is established;

[0007] (2) Using hierarchical coding method to code the hierarchical hexagonal grid after the space partition unit, record the hierarchical membership relationship between the hexagonal grid unit after hierarchical partition, form the spatial hierarchical constraint;

[0008] (3) Using Gosper space filling coding method to code the same spatial hierarchical hexagonal grid, recording the spatial proximity relationship of hexagonal grid unit on the same spatial hierarchy;

[0009] (4) Constructing spatial hierarchical network, the spatial partition level determines the hierarchy of spatial hierarchical network, and each level network is determined by the connection relationship between the space partition units. The hierarchical coding of hexagonal grid controls the spatial hierarchical relationship between the nodes of different hierarchical networks, and the Gosper space filling coding controls the spatial proximity relationship between the nodes of the same hierarchical network.

[0010] (5) For each hierarchical network, the heat map visualization method of network adjacency matrix can intuitively express the community structure characteristics of spatial network. Before visualization, the row and column nodes of the adjacency matrix are sorted from small to large using Gosper coding;

[0011] (6) For multiple network adjacency matrices of spatial hierarchical network, extract community block structure from top to bottom, and use the hierarchical constraint of the upper level to the next level when extracting community block structure of adjacent level network adjacency matrix, to mine hierarchical spatial community structure.

[0012] Further, in the step (1), the hierarchical hexagonal grid with an aperture of 7 is used to organize the hexagonal grid units after partition, and the hierarchical nesting relationship between different hierarchical hexagonal grids is established.

[0013] Further, in the step (2), the multi-bit hierarchical coding related to the partition level is used to code the hexagonal grid of different levels. The number of bits owned by the coding represents the level it is in. The first bit of the coding represents the initial partition level of the space partition unit, the second bit represents the first partition level, the third bit represents the second partition level, and the nth bit represents the n-1 partition level.

[0014] Further, in the step (3), the rule of filling the hexagonal grid space with Gosper curve is used to code the hexagonal grid of the same level according to the filling order, and the spatial proximity is maintained.

[0015] Further, in the step (4), in the same level hexagonal grid, the hexagonal space partition unit is taken as the network node, the network edge is established according to the connection relationship between the hexagonal grids, a spatial network is constructed in each level; the spatial hierarchical relationship is obtained between the spatial networks of adjacent levels through hierarchical coding, the spatial adjacent relationship is obtained between the nodes of the same level spatial network through the Gosper space filling coding, and the connection relationship between the hexagonal grids is determined by the actual network.

[0016] Further, in the step (5), the spatial hierarchical network is organized by using a plurality of network adjacency matrices, the row and column nodes of the adjacency matrix are sorted from small to large by using the Gosper coding, and the community structure characteristics of the spatial network are intuitively expressed in the form of a heat map visualization.

[0017] Further, in the step (6), for the heat map of the plurality of network adjacency matrices of the hierarchical spatial network, the block structure on the diagonal line from the last level network to the next level network is extracted as the community of the network, the extraction of the community of the next level network is hierarchically constrained by using the community of the last level network, and the hierarchical spatial community structure is obtained.

[0018] Compared with the prior art, the present application has the following advantages:

[0019] (1) The hierarchical spatial community structure obtained by the present application reflects the hierarchical block structure relationship of the spatial network phenomenon in the spatial region layout, the detection result reflects the spatial distribution of the spatial network topology, better reflects the characteristics of the real world multi-level (scale), and can better assist people to observe and study the geographical objects from the multi-level (scale) perspective.

[0020] (2) The present application adopts the Gosper-island (aperture 7) to hierarchically nest the partitioned hexagonal grid, can maximally cover the initial hexagonal grid partition unit, and is convenient for recording the hierarchical membership relationship of the spatial region. The hierarchical coding and the Gosper space filling coding are used to code the spatial partition unit, the hierarchical hexagonal grid spatial hierarchical relationship is recorded in the spatial longitudinal direction (vertical direction), and the spatial adjacent relationship of the hexagonal grid in the same spatial level is recorded in the spatial transverse direction.

[0021] (3) The present application adopts the top-down block structure extraction method, uses the community detection structure of the last level to constrain the community detection of the next level, conforms to the hierarchical characteristics of the hierarchical spatial community, and effectively avoids the problems caused by the hierarchical tree. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 The flowchart of the present application.

[0023] Figure 2Gosper-island hexagonal grid partitioning diagram in the application.

[0024] Figure 3 Gosper-island hierarchical partitioning diagram in the application.

[0025] Fig. 4(a), Fig. 4(b) and Fig. 4(c) are Gosper-island hierarchical encoding diagrams in the application, wherein Fig. 4(a) is an initial partitioning diagram, Fig. 4(b) is a first-level partitioning diagram, and Fig. 4(c) is a second-level partitioning diagram.

[0026] Figure 5 Gosper space-filling curve diagram in the application.

[0027] Figure 6 Gosper encoding diagram in the application.

[0028] Fig. 7(a) and Fig. 7(b) are top-down block structure extraction diagrams in the application, wherein Fig. 7(a) is a partial screenshot of an upper-level adjacency matrix, and Fig. 7(b) is a partial screenshot of a current-level adjacency matrix.

[0029] Figure 8 Square window diagram for block structure extraction in the application. DETAILED DESCRIPTION

[0030] The technical solutions of the application will be described in detail below with reference to the accompanying drawings.

[0031] A hierarchical spatial community detection method based on Gosper-island, first, the space is hierarchically partitioned by using Gosper-island (a hierarchical hexagonal grid with an aperture of 7), the spatial hierarchical constraints of the partitioned units are recorded using hierarchical partitioning encoding, and the spatial proximity constraints of the same hierarchical partitioned units are recorded using Gosper space-filling encoding; then, the partitioned units are used as network nodes to construct a hierarchical spatial network, and the network is organized in the form of a network adjacency matrix, the network adjacency matrix heat map is drawn according to the Gosper encoding order of the nodes; finally, the hierarchical matrix block structure is extracted from top to bottom for the network heat maps of different levels to obtain the hierarchical spatial community structure that meets the spatial hierarchical constraints.

[0032] Specifically, the method comprises the following steps:

[0033] (1) The space region is partitioned and hierarchically organized by using the Gosper-island hierarchical partitioning method, and the hierarchical nesting relationship between the hierarchical hexagonal grids in the space region is established;

[0034] (2) Adopt hierarchical coding method to code the space partition unit after hierarchical hexagonal grid partition, record the hierarchical membership relationship between the hexagonal grid units after hierarchical partition, and form the spatial hierarchical constraint;

[0035] (3) Adopt Gosper space filling coding method to code the same spatial hierarchical hexagonal grid, record the spatial proximity relationship of the hexagonal grid units on the same spatial hierarchy;

[0036] (4) Construct a spatial hierarchical network, the spatial partition level determines the hierarchy of the spatial hierarchical network, and each level of the network is determined by the connection relationship between the spatial partition units. The hierarchical coding of the hexagonal grid controls the spatial hierarchical relationship between the nodes of different hierarchical networks, and the Gosper space filling coding controls the spatial proximity relationship between the nodes of the same hierarchical network;

[0037] (5) For each hierarchical network, use the heat map visualization method of network adjacency matrix to intuitively express the community structure characteristics of the spatial network. Before visualization, sort the row and column nodes of the adjacency matrix from small to large using Gosper coding;

[0038] (6) For multiple network adjacency matrices of the spatial hierarchical network, extract the community block structure from top to bottom. When extracting the community block structure of the adjacent level network adjacency matrix, use the hierarchical constraint of the upper level to the lower level to mine the hierarchical spatial community structure.

[0039] In step (1), a regular hexagon is used to partition the space, and a Gosper-island (a hierarchical hexagonal grid with a hole diameter of 7) is used to organize the hexagonal grid units after partitioning, establishing the hierarchical nesting relationship between different hierarchical hexagonal grids.

[0040] In step (2), a multi-bit hierarchical coding related to the partition level is used to code the hexagonal grid of different levels. The number of bits owned by the coding represents the level it is in. The first bit of the coding represents the initial partition level of the space partition unit, the second bit represents the first partition level, the third bit represents the second partition level, and so on. All level attributions can be obtained in this way.

[0041] In step (3), according to the rules of Gosper curve space filling, the hexagonal grid is coded in the order of Gosper space filling, which records the spatial proximity between hexagonal grids to a certain extent.

[0042] In step (4), in the same level of hexagonal grid, the regular hexagon space partition unit is taken as the network node, and the network edge is established according to the connection relationship between the regular hexagons. In this way, a spatial network is constructed for each level. The spatial hierarchical relationship between the spatial networks of adjacent levels is obtained through hierarchical coding, and the spatial proximity relationship of the spatial network of the same level is obtained through the Gosper space filling coding between nodes. The connection relationship between the hexagonal grids is determined by the actual network connection.

[0043] In step (5), the spatial hierarchical network is organized by using multiple network adjacency matrices, the row and column nodes of the adjacency matrix are sorted from small to large using Gosper coding, and the community structure characteristics of the spatial network are intuitively expressed in the form of heat map visualization.

[0044] In step (6), for the heat map of the multiple network adjacency matrices of the hierarchical spatial network, the block structure on the diagonal line from the last level network to the next level network is extracted as the community of the network, and the hierarchical constraint of the extraction of the community of the last level network to the community of the next level network is used, so that the hierarchical spatial community structure is obtained.

[0045] The following is a more specific example:

[0046] As shown in Figure 1 , a hierarchical spatial community detection method based on Gosper-island includes the following steps:

[0047] First, the space region is partitioned by using regular hexagons, the partitioned hexagonal grid units are hierarchically organized by using Gosper-island, and the hierarchical nesting relationship between the hierarchical hexagonal grids in the space region is established. The isotropy, field consistency and high spatial coverage of the regular hexagon make it easy to be hierarchically partitioned, and the spatial characteristics of the partitioned units can be maximally maintained. The Gosper-island is a hierarchical hexagonal grid with a space of 7, which can maximally cover the initial hexagon by complete aggregation, and the Gosper-island hexagonal grid aggregation is as shown in Figure 2 . The Gosper-island hierarchical partitioning nesting process is as shown in Figure 3As shown, first, the initial hexagonal grid is spatially overlaid with the spatial region range to determine the initial hexagonal grid range covered by the spatial region range. On this basis, the initial hexagonal grid range is Gosper-island hierarchically nested to establish the nested relationship of the hierarchical hexagonal grid. Finally, according to the actual situation, the size of the hexagonal grid subdivision unit and the Gosper-island level are determined, and the hexagonal grid subdivision unit is used to subdivide the spatial region. The initial subdivision uses a complete Gopser-island to approximate the spatial range to be studied, and the best Gopser-island position and direction are found. After determining the initial Gosper-island, the subsequent iterative nesting is based on the Gosper-island, which continuously refines the hexagonal grid unit until the subdivision hexagonal grid unit size that meets the research requirements is obtained. Using the Gosper-island as the initial hexagonal grid for spatial overlay facilitates the next step of Gopser coding for the hexagonal grid hierarchical grid, simplifies the initial position addressing optimization process (NP infinite problem) of Gopser coding, reduces the complexity of the method, and thus saves the program operation time.

[0048] In the second step, the hierarchical coding method is used to organize the Gosper-island hierarchical hexagonal grid to obtain the coding of the hexagonal grid subdivision unit, record the hierarchical membership relationship between the hexagonal grid units after the subdivision, and form the spatial hierarchical constraint. The Gosper-island hierarchical coding is shown in FIG. 4, which uses a one-digit code to represent the seven hexagons of the initial Gosper-island unit, a two-digit code to represent the 49 hexagons of the first-level unit, a three-digit code to represent the 343 hexagons of the second-level unit, and so on. The hierarchical coding records the hierarchical membership relationship between different levels of hexagonal grids.

[0049] In the third step, the Gosper space filling coding method is used to code the same spatial hierarchical hexagonal grid to record the spatial proximity relationship of the hexagonal grid units at the same spatial level. The Gosper curve is used to fill the hexagonal grid space according to the filling order to code the hexagonal grid. The drawing rule of the Gosper curve is to replace each other with two groups of operations under the control of the iteration number. In order to describe the drawing rule of the Gosper curve clearly, the code is used to represent the combination of operations, as shown in the following table:

[0050] Table 1 Gosper curve drawing rule code

[0051]

[0052] Wherein, A and B represent fixed length of curve unit, generally A and B are equal in length. In the process of drawing, the rules are to replace A in the code with A replace operation code and replace B with B replace operation code. "--" represents a counterclockwise rotation of 60 degrees operation, "+" represents a clockwise rotation of 60 degrees operation. According to the requirement of the number of iterations, the replacement operation can complete the Gosper curve drawing. This method uses a certain edge vertex of Gosper-island as the starting point of curve drawing, and the length of the curve unit is six times the length of the hexagonal grid, as shown in the following figure. Figure 5 The use of such initial drawing parameter settings avoids the time-consuming starting point addressing operation when drawing at different levels, and the same starting point is used for Gosper curve drawing between different levels of the same Gosper-island.

[0053] The hexagonal grid unit is encoded in the process of Gosper curve drawing by using direct encoding method, and the encoding is the drawing order value of Gosper curve. The Gosper space-filling curve encoding is shown in the following figure. Figure 6

[0054] Fourthly, the space hierarchical network is constructed. The space hierarchical network is a multi-level space network set, each level of which corresponds to the subdivision level of the hierarchical nested model. There is a space hierarchy constraint between the space networks of adjacent levels, and the network nodes of adjacent levels have a complete containment relationship in space. The network node of each level network of the space hierarchical network is the space range of the bottommost hexagonal grid controlled by the corresponding hierarchical nested model nested unit. After the hierarchical nested model is used to subdivide the nested space region, the bottommost space subdivision unit (the smallest hexagonal grid) is first connected in space according to the actual network connection, the hierarchical coding of the smallest hexagonal grid is determined, and the smallest hexagonal grid is used as the network node to establish the bottommost space network. Then, according to the hierarchical coding of the current smallest hexagonal grid, the smallest hexagonal grid is merged into the second-to-last level hexagonal grid, and the second-to-last level hexagonal grid is used as the network node to re-statistic the connection relationship between the hexagonal grids, and the second-to-last level space network is constructed, and the construction of the space hierarchical network is gradually completed.

[0055] Fifthly, for each hierarchical network, the community structure characteristics of the space network are intuitively expressed by using the heat map visualization method of the network adjacency matrix. Before visualization, the row and column nodes of the adjacency matrix are sorted from small to large using Gosper coding. The network adjacency matrix heat map visualization is shown in the following figure.

[0056] Before observing and detecting the community structure of the network adjacency matrix, the weight values of the network adjacency matrix are standardized. The standardization function used is as follows.

[0057] ​​

[0058] Figure 7 is the adjacency matrix visualization result of the 2-level partition network of a certain hierarchical spatial network. After sorting the rows and columns of the network adjacency matrix using Gosper coding, the adjacency matrix is visualized by heat map, and it can be found that there is a more obvious block structure on the diagonal of the matrix.

[0059] In the sixth step, the community block structure is extracted from top to bottom for multiple network adjacency matrices of the spatial hierarchical network. When extracting the community block structure of the adjacent level network adjacency matrix, the hierarchical constraint of the upper level to the next level is used to mine the hierarchical community structure. The process of extracting the community block structure from top to bottom is shown in Figure 7, and the specific process is as follows:

[0060] First, the block structure of the upper network adjacency matrix is extracted, and the row and column nodes are identified by the community, to obtain the community structure of this layer. Then, the community division of the current level network adjacency matrix is performed, which needs to use the community division identification of the upper level as the input of the block structure extraction of the current level network adjacency matrix. According to the community division identification of the upper level, each community node is extracted, the hierarchical coding of each community node is queried, the hierarchical coding of the current level network node controlled by the community node hierarchical coding is queried, it is converted into Gosper coding, and the adjacency matrix expression is constructed according to the Gosper coding sorting, as shown in Figure 7(b). Finally, each community block structure is traversed, and the block structure of the current level adjacency matrix is extracted under the control of the block structure. The block structure of each layer adjacency matrix of the hierarchical network is extracted in this way until the bottom layer network is reached, or the community division requirement is met.

[0061] In this method, the extraction of the block structure can be completed by means of the related matrix calculation operation, which is divided into two steps:

[0062] First, the network adjacency matrix is converted into a TF matrix using a threshold, and the block structure is roughly extracted. The TF matrix (TFM) is a matrix with element values of "True" or "False". TFM is a function transformation of the adjacency matrix M, TFM = {tfm ij} is the structure matrix of the adjacency matrix, and the matrix value element of the structure matrix is "True" or "False".

[0063]

[0064] Where M = {m ij} is the network adjacency matrix, m ij is the normalized weight value of the adjacency matrix, m α is the network weight threshold, and the elements greater than the threshold are "True" in the corresponding row and column position of the TFM matrix, m αThe setting needs to be combined with specific application examples to set the threshold value to extract the block structure on the diagonal line of the adjacency matrix in coarse granularity. In the hierarchical spatial network experiment of the method, setting m α =2, the structure matrix of the adjacency matrix can be extracted more clearly.

[0065] Secondly, a square window is set based on the diagonal line as the reference axis, the side length of the square is continuously enlarged, and the row and column unit of the adjacency matrix is increased each time, as shown in the following formula: Figure 8 The red frame line represents the structure matrix, and the black frame line represents the square window to be set.

[0066] The top-left point of the square window is the element with the row and column code (i, i) in the structure matrix, and the bottom-right point is the element with the row and column code (j, j) in the structure matrix. The value of B ratio (i, j) in the window is calculated.

[0067]

[0068] num(TFM i→j , "True") is the number of "True" element values in the square window, and num(i→j) is the number of element values in the square window, which is 1, 4, 9, 16, and the like.

[0069] When , the side length of the square is increased by 1 unit. If the condition is not met, the i→(j-1) row and column of the structure matrix are extracted as the block structure, the community identifier is set, the starting point coordinates of the square window are reset as i=j and j=j+1, and the second step is repeated until the square window covers the diagonal line of the structure matrix.

[0070] In summary, the block structure extraction method from top to bottom is adopted, the community detection structure of the upper level is used to constrain the community detection of the next level, which meets the hierarchical characteristics of the hierarchical spatial community, can effectively avoid the problems caused by the partition hierarchical tree, and the obtained hierarchical spatial community structure reflects the hierarchical block structure relationship of the spatial network phenomenon in the spatial region layout, better reflects the multi-level (scale) characteristics of the real world, and can better assist people to observe and study geographical objects from the multi-level (scale) perspective.

Claims

1. A Gosper-island-based hierarchical spatial community mining method, characterized in that, The method comprises the following steps: (1) using regular hexagons to divide a space region, using Gosper-island to organize the hexagonal grid cells after division, and establishing a hierarchical nesting relationship between the hexagonal grid cells in the space region; (2) using a hierarchical coding method to code the space division cells after the hierarchical hexagonal grid division, recording the hierarchical membership relationship between the hexagonal grid cells after the hierarchical division, and forming a spatial hierarchical constraint; (3) using a Gosper space filling coding method to code the same space hierarchical hexagonal grid, and recording the spatial neighborhood relationship of the hexagonal grid cells on the same space hierarchy; (4) constructing a space hierarchical network, the space division level determining the hierarchy of the space hierarchical network, each level of the network being determined by the connection relationship between the space division cells, the hierarchical coding of the hexagonal grid controlling the spatial hierarchical relationship between the nodes of different levels of the network, and the Gosper space filling coding controlling the spatial neighborhood relationship between the nodes of the same level of the network; (5) for each level of the network, using a heat map visualization method of a network adjacency matrix to intuitively express the community structure characteristics of the space network, and sorting the row and column nodes of the adjacency matrix from small to large using the Gosper coding before visualization; (6) for a plurality of network adjacency matrices of the space hierarchical network, extracting community block structures from top to bottom, and using the hierarchical constraint of the upper level on the lower level when extracting the community block structures of the network adjacency matrices of the adjacent levels, to mine hierarchical space community structures. 2.The hierarchical spatial community mining method based on Gosper-island according to claim 1, wherein, In the step (1), the hierarchical hexagonal grid with an aperture of 7 is used to organize the hexagonal grid cells after division, and the hierarchical nesting relationship between the hexagonal grid cells at different levels is established. 3.The hierarchical spatial community mining method based on Gosper-island according to claim 1, wherein, In the step (2), a multi-bit hierarchical coding related to the division level is used to code the hexagonal grid at different levels, the number of bits of the coding representing the level at which the coding is located, the first bit of the coding representing the initial division level of the space division cell, the second bit representing the first division level, the third bit representing the second division level, and the nth bit representing the n-1 division level.

4. The hierarchical spatial community mining method based on Gosper-island of claim 1, wherein, In the step (3), the Gosper curve is used to fill the hexagonal grid space according to the filling order, and the same level of the hexagonal grid is coded, so that the spatial neighborhood is maintained.

5. The hierarchical spatial community mining method based on Gosper-island according to claim 1, wherein, In the step (4), in the same level of the hexagonal grid, the regular hexagonal space division cell is used as a network node, and a network edge is established according to the connection relationship between the regular hexagons, and a space network is constructed at each level; the space hierarchical relationship between the space networks of adjacent levels is obtained through the hierarchical coding, the space neighborhood relationship between the nodes of the same level of the space network is obtained through the Gosper space filling coding between the nodes, and the connection relationship between the hexagons is determined by the actual connection of the actual network.

6. The hierarchical spatial community mining method based on Gosper-island of claim 1, wherein, In the step (5), a plurality of network adjacency matrices are used to organize the space hierarchical network, the row and column nodes of the adjacency matrix are sorted from small to large using the Gosper coding, and a heat map visualization method is used to intuitively express the community structure characteristics of the space network.

7. The hierarchical spatial community mining method based on Gosper-island of claim 1, wherein, In the step (6), for the plurality of network adjacency matrix heat maps of the hierarchical spatial network, the block structure on the diagonal line from the previous level network to the next level network is extracted, that is, the community of the network, and the extraction of the community of the next level network is hierarchically constrained by the community of the previous level network, to obtain the hierarchical spatial community structure.

Images