Support form prediction method and system based on multi-scale density peak clustering

Abstract

The invention discloses a support form prediction method and system based on multi-scale density peak clustering, and belongs to the field of information science. Comprising the following steps: firstly, calculating the similarity of different coal mine multi-scale support forms and the semantic similarity of various attributes influencing support form decisions by using coal mine support parameter data acquired on site; then, constructing a hypergraph based on the multi-scale support form similarity and the roadway attribute semantic similarity to represent roadway attributes; thirdly, using an attribute density peak mechanism and a community detection method on the roadway attribute hypergraph to identify an attribute hotspot overlapping community; and finally, analyzing the distribution condition of the given attribute in the community, and completing support form prediction.

Description

Technical Field

[0001] The present invention belongs to the field of information science, and specifically relates to a support form prediction method and system based on multi-scale density peak clustering. Background Art

[0002] Coal mine support is to support and reinforce geological structures such as coal mine working faces and tunnels to ensure the stability and safety of the working environment. In the past, the research on support effect mainly used numerical simulation software to analyze the anchor support of coal tunnels, obtain the stress distribution of the surrounding rock of the tunnel anchor support and the law of mine pressure manifestation, to verify the rationality of the support, and to provide a theoretical basis for optimizing the anchor support parameters of coal tunnels and controlling the deformation of surrounding rocks. There are few reports on the use of artificial intelligence prediction methods for the design of support forms.

[0003] The intelligent coal mine support refers to the use of basic parameters that are easy to measure, such as coal mine tunnel characteristics and surrounding rock characteristics, and the support form prediction method based on multi-scale density peak clustering of the present invention to predict the diameter, length, spacing and other parameters of the roof anchor, side anchor and anchor cable, and finally design a feasible support plan based on the prediction results.

[0004] The density peak clustering algorithm can effectively identify clusters of arbitrary shapes, is suitable for large-scale data sets and does not require a preset number of clusters. It is simple and efficient and is widely used in cluster analysis tasks in various fields. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention proposes a support form prediction method and system based on multi-scale density peak clustering to predict the diameter, length, spacing and other parameters of the top plate anchor rods, side anchor rods and anchor cables, and designs a feasible support scheme according to the prediction results.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A support form prediction method based on multi-scale density peak clustering includes the following steps:

[0008] Using the coal mine support parameter data collected in the field, the similarity of multi-scale support forms of different coal mines and the semantic similarity of various tunnel attributes that affect the support form decision are calculated;

[0009] Based on the similarity of the multi-scale support forms of different coal mines and the semantic similarity of the roadway attributes, a hypergraph of roadway attributes-support forms is constructed;

[0010] On the hypergraph of lane attribute-support form, the lane attribute density peak mechanism and community detection method are applied to identify the overlapping communities of attribute hot spots;

[0011] Analyze the distribution of given attributes in the attribute hotspot overlapping communities to complete the support form prediction. Preferably, the coal mine support parameter data is expressed as:

[0012] S i ={(t1,W 1i ),(t2,W 2i ),...,(t n ,W ni )},

[0013] Among them, W ni Represents attribute t n In the feature set S i The relative importance of .

[0014] Preferably, the method for calculating the similarity of multi-scale support forms of different coal mines includes:

[0015]

[0016] Among them, l i , l j All are lanes, Cid t (l i ) represents the attribute t for lane l i The support form influence score, Cid t (l j ) represents the attribute t for lane l j The support form influences the score.

[0017] Preferably, the method for calculating the semantic similarity of multiple tunnel attributes that affect support form decision-making includes:

[0018] l i and l j are in the same connected component of graph G, which is a directed acyclic graph, that is, there is no t i and t j If there is a circular reference between i and l j The semantic similarity is defined as:

[0019]

[0020] l i and l j In the same connected component of graph G, t i and t j is on a ring in the connected component, that is, there exists t i and t j If there is a circular reference between i and l j The semantic similarity is defined as:

[0021]

[0022] l i and l j In different connected components of graph G, that is, t i and t j If there is no explicit association between i and l j The semantic similarity is defined as:

[0023]

[0024] Preferably, the method for constructing a hypergraph of tunnel attributes and support forms includes:

[0025] Use G = (L, E, W) to represent a hypergraph, where L is the set of lanes, that is, the set of vertices of the hypergraph G; E is the set of edges, representing any two points l i and l j There is an equal-value association between them or an association on the attribute t∈T, where T represents the attribute set; W is the set of weights, indicating the degree of association of E.

[0026] Preferably, the method of applying the lane attribute density peak mechanism and the community detection method to identify attribute hotspot overlapping communities includes:

[0027] Calculate lane l i The local density ρ(l i );

[0028] Based on the local density ρ(l i ), calculate the average density on the hypergraph G of tunnel attributes and support forms;

[0029] Calculate lane l i The density deviation degree δ(l i );

[0030] Based on the average density of the hypergraph G of the tunnel attribute-support form and the tunnel l i The density deviation degree is obtained by local density ρ(l i ) is the x-axis, and the density deviation degree δ(l i ) is the decision diagram with y axis;

[0031] Based on the decision graph, obtaining diversified density peak nodes;

[0032] Initialize the density peak node as the initial community core, and use the existing density peak clustering algorithm to divide the hypergraph G of the lane attribute-support form into a number of dense subgraphs;

[0033] The algorithm stops when all lanes are assigned to at least one community;

[0034] For the community core set C = {c1, c2, ..., c k}, and the detected overlapping communities G = {g1,,g2,…,g k}.

[0035] Preferably, the method of analyzing the distribution of a given attribute in the overlapping communities of attribute hot spots and completing the support form prediction includes:

[0036] Using a probability vector P i =[p i1 ,p i2 ,...,p in ], indicating lane l i The probability of belonging to each community,

[0037] For the community j ,have:

[0038] Among them, A + (l i ) is composed of l i and the set of adjacent nodes, d u (l i , l k ) represents lane l i , l k The result of normalizing the attribute deviation degree, d u (l x , l y ) represents lane l x , l y The result of normalizing the degree of attribute deviation;

[0039] Using vector C t =[c t1 , c t2 , ..., x tn ] represents the influence of attribute t, where

[0040] For lane l, it belongs to every community in G, and attribute t is also distributed in every community. The influence score of attribute t in G is calculated as

[0041] Calculate the top-k attributes of the influential scores of lane l. Then, among the known lane attributes, the support form with the highest frequency that has the same top-k attributes is the prediction of the support form of lane l.

[0042] The present invention also discloses a support form prediction system based on multi-scale density peak clustering, comprising: a calculation module, a construction module, an identification module and a prediction module;

[0043] The calculation module is used to calculate the similarity of multi-scale support forms of different coal mines and the semantic similarity of various tunnel attributes that affect the decision of support form by using the coal mine support parameter data collected on site;

[0044] The construction module is used to construct a hypergraph of tunnel attribute-support form based on the similarity of multi-scale support forms of different coal mines and the semantic similarity of tunnel attributes;

[0045] The identification module is used to apply the lane attribute density peak mechanism and community detection method on the hypergraph of lane attribute-support form to identify the attribute hotspot overlapping communities;

[0046] The prediction module is used to analyze the distribution of a given attribute in the overlapping communities of the attribute hotspots and complete the support form prediction.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The multi-scale density peak clustering algorithm can more accurately identify and classify different features and patterns in tunnel support. By analyzing data at different scales, more details and complex structures can be captured, improving the accuracy of support prediction. This multi-scale analysis method can better adapt to complex tunnel environments and effectively improve the reliability and safety of support solutions.

[0049] 2. The multi-scale density peak clustering algorithm is highly adaptable and robust, and can handle data of different types and sizes. Whether it is large-scale data or noisy data, the algorithm can effectively identify key patterns and trends in the data. This is particularly important for tunnel support prediction, because the tunnel environment is complex and changeable, and the data quality is uneven. The robustness of the algorithm makes it perform well in practical applications and can provide reliable prediction results in various situations.

[0050] 3. The algorithm determines the cluster center by effectively finding the density peak, avoiding a large number of iterations and complex calculation processes, thereby greatly reducing the calculation complexity and time cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0052] Figure 1 The present invention is a flowchart of a method for predicting support forms based on multi-scale density peak clustering according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] Embodiment 1

[0056] like Figure 1 As shown, the present invention provides a support form prediction method based on multi-scale density peak clustering, comprising the following steps:

[0057] Using the coal mine support parameter data collected in the field, the similarity of multi-scale support forms of different coal mines and the semantic similarity of various tunnel attributes that affect the support form decision are calculated;

[0058] Based on the similarity of the multi-scale support forms of different coal mines and the semantic similarity of the roadway attributes, a hypergraph of roadway attributes-support forms is constructed;

[0059] On the hypergraph of lane attribute-support form, the lane attribute density peak mechanism and community detection method are applied to identify the overlapping communities of attribute hot spots;

[0060] The distribution of a given attribute in the overlapping communities of the attribute hotspots is analyzed to complete the support form prediction.

[0061] After preliminary on-site investigation, field survey, and interview collection, the present invention stores the final 295 data in the SQLite database. The fields include: coal seam thickness, direct top lithology, direct top thickness, old top lithology, old top thickness, direct bottom lithology, direct bottom thickness, old bottom lithology, old bottom thickness, coal pillar width, cross-sectional shape, tunnel burial depth, and support form. Except for the last support form, the rest are input parameters of the prediction method, i.e., tunnel attribute characteristics. After eliminating samples with missing values, 270 data are obtained.

[0062] In the intelligent support system of coal mines, the coal mines that need to be predicted for support form usually include a variety of different attributes (such as geographical location expressed in longitude and latitude, direct roof lithology, direct roof thickness, etc., the total number of attributes ranges from dozens to dozens), and these attributes jointly affect the support form. The characteristics of the coal mine are characterized by the tunnel attribute features (attribute weight scheme). By extracting these attributes, the feature set of a tunnel is represented as S i ={(t1, W 1i ), (t2, W 2i ),...,(t n , W ni )}, which is a key-value pair set of attribute-weight value. xi Represents attribute t x In the feature set S i The relative importance in is calculated using the term frequency-inverse document (TF-IDF) algorithm. The specific calculation method is as follows:

[0063]

[0064] Among them, freq xi is the attribute t x The word frequency of attribute-attribute value (attribute value is a set) is x =(v1, v2, ..., v m ),freq xi = m, which is numerically equal to the size of the attribute value set; max l (freq li ) indicates S i The word frequency of the attribute with the maximum word frequency in N A is the total number of records of coal mine support forms, n x It contains the attribute t x The weight can describe the importance of the roadway attribute for the prediction of the roadway support form.

[0065] Furthermore, all possible attributes t are represented as a set T, and the content impact degree (CID) of attribute t in the support data of any roadway is calculated using the following formula:

[0066]

[0067] In which, when t∈T, φ(t, T)=1, otherwise it is 0.

[0068] The above analysis can be used to calculate the similarity of multi-scale support forms of different coal mine roadways, as well as to calculate the attribute semantic similarity of various attributes that affect the decision of roadway support forms.

[0069] If the attribute t content of two tunnels is exactly the same, then their support form similarity should be greater than those of completely different tunnels, so the following formula is used to model the multi-scale support form similarity (Content Similarity, CSim) of different coal mine tunnels:

[0070]

[0071] The above formula uses the attribute t for lane l i and Lane j The support form influence degree is taken into consideration, and the support form similarity of the two tunnels is calculated, where i and j represent the numbers of different tunnels. Since attribute t comes from the attribute set T, the calculation process involving CSim of multiple attributes is the calculation of multi-scale support form similarity.

[0072] There is a semantic association between the attribute values ​​of the lanes. Specifically, there is an attribute t a , the range of its attribute value is where n a Represents attribute t a The number of features included; the existence of attribute t b , the range of its attribute value is where n b Represents attribute t b The number of features included. For lane i, For t b , there is a certain set of attribute values Make Attributes a and t b Whether there is an association and the constraints on their respective value ranges are prior knowledge. i and l j , if lane l i and l j If the same constraints appear in their respective attributes, they are considered to be related and there is an edge between them. <l i , l j >, based on this we can get graph G L Then for l i and l j In Figure G L There are three semantic similarity (SSim) calculations for different situations:

[0073] l i and l j In the same connected component of graph G, this connected component is a directed acyclic graph, that is, there is no t i and tj If there is a circular reference between i and l j The semantic similarity is defined as:

[0074]

[0075] l i and l j In the same connected component of graph G, t i and t j is on a ring in this connected component, that is, there exists t i and t j If there is a circular reference between i and l j The semantic similarity is defined as:

[0076]

[0077] l i and l j In different connected components of graph G, that is, t i and t j If there is no explicit association between i and l j The semantic similarity is defined as:

[0078]

[0079] Finally, for attribute t, the similarity between tunnel support attributes is obtained by weighting the multi-scale support form similarity and semantic similarity:

[0080] Sim t (l i , l j )=αSSim t (l i , l j )+(1-α)CSim t (l i , l j ),

[0081] Among them, α is a hyperparameter that reflects the relative importance of the single attribute of the roadway and the semantic association between attributes to the support form.

[0082] If any two lanes have the same attribute value, they are considered to have an equal value association. In order to calculate this association, for lane l i and l j , use E i and E j Respectively represent the lane l i and l jThere is a collection of associated lanes, lane l i and l j The degree of equivalence association is:

[0083]

[0084] Use G = (L, E, W) to represent a hypergraph, where: L is the set of lanes, that is, the set of vertices of the hypergraph G; E is the set of edges, representing any two points l i and l j There is an equal association between them or an association between attributes t∈T; W is a set of weights, indicating the degree of association of E, that is, W={Sim t |t∈T}∪{Sim E}. Use d(l)=Σ e∈E h(l, e) represents the degree of lane l. If l is an endpoint of edge e, h(l, e) = 1, otherwise it is 0.

[0085] Based on the above definition of hypergraph and association degree, the lane l is given i , l j The degree of deviation of the attribute Calculation:

[0086]

[0087] Degree of deviation from attribute After normalization, we get d u (l i , l j ):

[0088]

[0089] Obviously there is u (l i ,l j ) = d u (l j ,l i ), that is, the association between lanes is mutual and undirected.

[0090] The number of connections between vertices is not uniform, reflecting the different degrees of association between different lanes. The density of each vertex is different, reflecting that the number of attributes owned by the lanes is not exactly the same. A density peak clustering method is designed to classify objects into local maxima of the density field. The method has a basic assumption that the cluster core is found by relatively high local link density and is relatively far away from any point with high local density. Use the existing density peak idea to find the core of the hypergraph G, and then perform overlapping community detection.

[0091] Lanei The local density ρ(l i ) is calculated as:

[0092]

[0093] Among them A + (l i )=A(l i )U{l i}, is composed of l i and the set of nodes adjacent to it. χ(·) is used for binarization:

[0094]

[0095] The average density on graph G is:

[0096]

[0097] l i The density deviation degree δ(l i ) is calculated as:

[0098]

[0099] Based on the definition of density and deviation, we can get a density ρ(l i ) is the x-axis, and the deviation degree δ(l i ) is the decision graph with y-axis. Nodes with higher local density ρ(·) and deviation degree δ(·) can be regarded as diversified density peak nodes. These density peak nodes are initialized as the initial community cores, and then the existing density peaks clustering (DPC) algorithm is used to divide the hypergraph into several robust dense subgraphs. When all vertices are assigned to at least one community, the algorithm stops. According to the above idea, for the community core set C = {c1, c2, ..., c k}, where k is the number of communities, c i Denotes the core of the ith community, we can obtain the monitored overlapping communities G = {g1, g2, ..., g k}. Through covering search, the nodes of each community either overlap with other communities or do not overlap. That is, lane l may belong to multiple communities, and two communities may have many overlapping lanes. The overlap coefficient of two communities is defined as γ ij :

[0100]

[0101] Then, the communities in G whose overlap coefficients are greater than the threshold η are continuously merged pairwise until the overlap coefficients between all communities in G are less than the threshold η.

[0102] After the community is merged, there are n≥1 communities, G={g1,g2,...,g n}, first, use a probability vector P i =[p i1 , p i2 , ..., p in ], indicating lane l i The probability of belonging to each community, where p ij Indicates lane l i The probability of belonging to the jth community. For community g j , p ij The calculation is as follows:

[0103] Among them, d u (l i , l k ) represents lane l i , l k The result of normalizing the attribute deviation degree, d u (l x , l y ) represents lane l x , l y The result of normalizing the degree of attribute deviation;

[0104] Then use the vector C t =[c t1 , c t2 , ..., c tn ] represents the influence of attribute t, where in Represents any element v in the community to the community g j Influence score.

[0105] For lane l, it may belong to every community in G, and attribute t is also distributed in every community. The influence score of attribute t in G is calculated as Calculate the top-k attributes of the influential scores of lane l. Then, among the known lane attributes, the support form with the highest frequency that has the same top-k attributes is the prediction of the support form of lane l.

[0106] Embodiment 2

[0107] The present invention also discloses a support form prediction system based on multi-scale density peak clustering, comprising: a calculation module, a construction module, an identification module and a prediction module;

[0108] The calculation module is used to calculate the similarity of multi-scale support forms of different coal mines and the semantic similarity of various tunnel attributes that affect the decision of support form by using the coal mine support parameter data collected in the field;

[0109] The construction module is used to construct a hypergraph of tunnel attribute-support form based on the similarity of multi-scale support forms of different coal mines and the semantic similarity of tunnel attributes;

[0110] The identification module is used to apply the lane attribute density peak mechanism and community detection method on the lane attribute-support form hypergraph to identify the attribute hotspot overlapping communities;

[0111] The prediction module is used to analyze the distribution of a given attribute in the overlapping communities of the attribute hotspots and complete the support form prediction.

[0112] In this embodiment, the coal mine support parameter data is expressed as:

[0113] S i ={(t1,W 1i ),(t2,W 2i ),...,(t n ,W ni )},

[0114] Among them, W ni Represents attribute t n In the feature set S i The relative importance of .

[0115] In this embodiment, the process of calculating the similarity of multi-scale support forms of different coal mines includes:

[0116]

[0117] Among them, l i , l j All are lanes, Cid t (l i ) represents the attribute t for lane l i The support form influence score, Cid t (l j ) represents the attribute t for lane l j The support form influences the score.

[0118] In this embodiment, the method for calculating the semantic similarity of multiple tunnel attributes that affect the support form decision includes:

[0119] l i and l j are in the same connected component of graph G, which is a directed acyclic graph, that is, there is no t i and t jIf there is a circular reference between i and l j The semantic similarity is defined as:

[0120]

[0121] l i and l j In the same connected component of graph G, t i and t j is on a ring in the connected component, that is, there exists t i and t j If there is a circular reference between i and l j The semantic similarity is defined as:

[0122]

[0123] l i and l j In different connected components of graph G, that is, t i and t j If there is no explicit association between i and l j The semantic similarity is defined as:

[0124]

[0125] In this embodiment, the method for constructing a hypergraph of tunnel attributes and support forms includes:

[0126] Use G = (L, E, W) to represent a hypergraph, where L is the set of lanes, that is, the set of vertices of the hypergraph G; E is the set of edges, representing any two points l i and l j There is an equal-value association between them or an association on the attribute t∈T, where T represents the attribute set; W is the set of weights, indicating the degree of association of E.

[0127] In this embodiment, the method of identifying attribute hotspot overlapping communities by applying the lane attribute density peak mechanism and the community detection method includes:

[0128] Calculate lane l i The local density ρ(l i );

[0129] Based on the local density ρ(l i ), calculate the average density on the hypergraph G of tunnel attributes and support forms;

[0130] Calculate lane l i The density deviation degree δ(l i );

[0131] Based on the average density of the hypergraph G of the tunnel attribute-support form and the tunnel l i The density deviation degree is obtained by local density ρ(l i ) is the x-axis, and the density deviation degree δ(l i ) is the decision diagram with y axis;

[0132] Based on the decision graph, obtaining diversified density peak nodes;

[0133] Initialize the density peak node as the initial community core, and use the existing density peak clustering algorithm to divide the hypergraph G of the lane attribute-support form into a number of dense subgraphs;

[0134] The algorithm stops when all lanes are assigned to at least one community;

[0135] For the community core set C = {c1, c2, ..., c k}, and the detected overlapping communities G = {g1,,g2,...,g k}.

[0136] In this embodiment, the method of analyzing the distribution of a given attribute in the overlapping communities of attribute hot spots and completing the support form prediction includes:

[0137] Using a probability vector P i =[p i1 , p i2 , ..., p in ], indicating lane l i The probability of belonging to each community,

[0138] For the community j ,have:

[0139] Among them, A + (l i ) is composed of l i and the set of adjacent nodes, d u (l i , l k ) represents lane l i , l k The result of normalizing the attribute deviation degree, d u (l x , l y ) represents lane l x , l y The result of normalizing the degree of attribute deviation;

[0140] Using vector C t=[c t1 , c t2 , ..., c tn ] represents the influence of attribute t, where

[0141] For lane l, it belongs to every community in G, and attribute t is also distributed in every community. The influence score of attribute t in G is calculated as

[0142] Calculate the top-k attributes of the influential scores of lane l. Then, among the known lane attributes, the support form with the highest frequency that has the same top-k attributes is the prediction of the support form of lane l.

[0143] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A support form prediction method based on multi-scale density peak clustering, characterized in that: The following steps are involved: Using the coal mine support parameter data collected in the field, the similarity of multi-scale support forms of different coal mines and the semantic similarity of various tunnel attributes that affect the support form decision are calculated; Based on the similarity of the multi-scale support forms of different coal mines and the semantic similarity of the roadway attributes, a hypergraph of roadway attributes-support forms is constructed; On the hypergraph of lane attribute-support form, the lane attribute density peak mechanism and community detection method are applied to identify the overlapping communities of attribute hot spots; The distribution of a given attribute in the overlapping communities of the attribute hotspots is analyzed to complete the support form prediction.

2. The support form prediction method based on multi-scale density peak clustering according to claim 1 is characterized in that: The coal mine support parameter data is expressed as: S i ={(t1,W 1i ),(t2,W 2i ),...,(t n ,W ni )}, Among them, W ni Represents attribute t n In the feature set S i The relative importance of .

3. The support form prediction method based on multi-scale density peak clustering according to claim 2 is characterized in that: The methods for calculating the similarity of multi-scale support forms in different coal mines include: Among them, l i , l j All are lanes, Cid t (l i ) represents the attribute t for lane l i The support form influence score, Cid t (l j ) represents the attribute t for lane l j The support form influences the score.

4. The support form prediction method based on multi-scale density peak clustering according to claim 3 is characterized in that: Methods for calculating the semantic similarity of various tunnel attributes that affect support form decisions include: l i and l j are in the same connected component of graph G, which is a directed acyclic graph, that is, there is no t i and t j If there is a circular reference between i and l j The semantic similarity is defined as: l i and l j In the same connected component of graph G, t i and t j is on a ring in the connected component, that is, there exists t i and t j If there is a circular reference between i and l j The semantic similarity is defined as: l i and l j In different connected components of graph G, that is, t i and t j If there is no explicit association between i and l j The semantic similarity is defined as:

5. The support form prediction method based on multi-scale density peak clustering according to claim 1 is characterized in that: Methods for constructing a hypergraph of tunnel attributes and support forms include: Use G = (L, E, W) to represent a hypergraph, where L is the set of lanes, that is, the set of vertices of the hypergraph G; E is the set of edges, representing any two points l i and l j There is an equal-value association between them or an association on the attribute t∈T, where T represents the attribute set; W is the set of weights, indicating the degree of association of E.

6. The support form prediction method based on multi-scale density peak clustering according to claim 1 is characterized in that: Methods for identifying overlapping communities of attribute hotspots by applying the lane attribute density peak mechanism and community detection methods include: Calculate lane l i The local density ρ(l i ); Based on the local density ρ(l i ), calculate the average density on the hypergraph G of tunnel attributes and support forms; Calculate lane l i The density deviation degree δ(l i ); Based on the average density of the hypergraph G of the tunnel attribute-support form and the density deviation degree of the tunnel li, a local density ρ(l i ) is the X-axis, and the density deviation degree δ(l i ) is the decision diagram with y axis; Based on the decision graph, obtaining diversified density peak nodes; Initialize the density peak node as the initial community core, and use the existing density peak clustering algorithm to divide the hypergraph G of the lane attribute-support form into a number of dense subgraphs; The algorithm stops when all lanes are assigned to at least one community; For the community core set C = {c1, c2, ..., c k }, and the detected overlapping communities G = {g1, g2, ..., g k }.

7. The support form prediction method based on multi-scale density peak clustering according to claim 1 is characterized in that: The method of analyzing the distribution of a given attribute in the overlapping communities of the attribute hot spots and completing the support form prediction includes: Using a probability vector P i =[p i1 , p i2 , ..., p in ], indicating lane l i The probability of belonging to each community, For the community j ,have: Among them, A+(l i ) is composed of l i and the set of adjacent nodes, d u (l i , l k ) represents lane l i , l k The result of normalizing the attribute deviation degree, d u (l x , l y ) represents lane l x , l y The result of normalizing the degree of attribute deviation; Using vector C t =[c t1 , c t2 , ..., c tn ] represents the influence of attribute t, where For lane l, it belongs to every community in G, and attribute t is also distributed in every community. The influence score of attribute t in G is calculated as Calculate the top-k attributes of the influential scores of lane l. Then, among the known lane attributes, the support form with the highest frequency that has the same top-k attributes is the prediction of the support form of lane l.

8. A support form prediction system based on multi-scale density peak clustering, characterized in that: include: Computation module, construction module, recognition module and prediction module; The calculation module is used to calculate the similarity of multi-scale support forms of different coal mines and the semantic similarity of various tunnel attributes that affect the decision of support form by using the coal mine support parameter data collected on site; The construction module is used to construct a hypergraph of tunnel attribute-support form based on the similarity of multi-scale support forms of different coal mines and the semantic similarity of tunnel attributes; The identification module is used to apply the lane attribute density peak mechanism and community detection method on the hypergraph of lane attribute-support form to identify the attribute hotspot overlapping communities; The prediction module is used to analyze the distribution of a given attribute in the overlapping communities of the attribute hotspots and complete the support form prediction.

Images