A traffic checkpoint importance ranking method based on word embedding vector and PageRank algorithm

Abstract

The present invention discloses a method for ranking the importance of traffic checkpoints based on embedded word vectors and PageRank algorithm, comprising the following steps: (1) obtaining a data source and preprocessing the data of the data source; (2) extracting a set of traffic checkpoints for representing locations; introducing a Glove model in natural language processing, and obtaining an embedded word vector representation of each checkpoint by training vehicle trajectory data; (3) ranking the importance of each checkpoint in combination with an improved PageRank algorithm; (4) using a node deletion method to simulate the situation where a checkpoint fails for some reason, and after removing the specified checkpoint, verifying the effectiveness of the importance ranking of each checkpoint in step (3) through prediction error. The present invention is aimed at ranking key points on roads, taking into account the physical topological structure of the traffic network, traffic demand and actual travel trajectory, and more objectively deriving the importance of different checkpoints from multiple perspectives; it can well take into account the influence of history, and can perform ranking algorithm verification.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic checkpoint importance ranking, and in particular to a traffic checkpoint importance ranking method based on embedded word vectors and PageRank algorithm. Background Art

[0002] Over the years, practical experience in transportation development both domestically and internationally has proven that solving urban traffic problems cannot rely solely on building basic road infrastructure; modern traffic management is also a crucial component. Therefore, establishing an intelligent, safe, efficient, convenient, comfortable, and environmentally friendly transportation system is particularly important.

[0003] From a macro perspective, understanding the correlation of traffic characteristics between nodes across the entire road network and identifying key nodes is fundamental to solving urban road traffic problems. Understanding the correlation of traffic characteristics between different nodes not only enables traffic flow prediction but also estimates network traffic demand. Identifying key nodes not only provides insights into the topological complexity of the road network but also helps traffic managers determine the structure of urban communities, facilitating traffic planning.

[0004] From a microscopic perspective, accurately and efficiently identifying and extracting key nodes within a city's complex transportation network can inform traffic managers' decision-making. Traveler information is often embedded in vehicle trajectory data. By extracting and analyzing vehicle trajectories, we can understand the travel habits and patterns of different travelers, thereby identifying potential points of interest in the city. This helps traffic managers provide travel advice to travelers, thereby avoiding traffic congestion, improving road network capacity, and achieving efficient operation of the entire transportation system.

[0005] However, existing algorithms for ranking checkpoint importance are primarily defined for static networks, whereas real-world transportation networks are dynamic, evolving over time. Furthermore, simply measuring the similarity between checkpoints using traditional graph theory methods fails to capture the underlying information, making it difficult to effectively identify key nodes in the road network and understand the topological characteristics of urban transportation. Summary of the Invention

[0006] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a method for ranking the importance of traffic checkpoints based on embedded word vectors and PageRank algorithm.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] The present invention provides a method for ranking traffic checkpoint importance based on word embedding vectors and PageRank algorithm, comprising the following steps:

[0009] (1) Obtain the data source including the user's vehicle information and the collection of vehicle passing information at each checkpoint, and pre-process the data from the data source to extract the user's vehicle trajectory;

[0010] (2) Extracting the set of traffic checkpoints used to represent the location in the user's vehicle trajectory in step (1); introducing the Glove model in natural language processing, and obtaining the embedded word vector representation of each checkpoint by training the vehicle trajectory data;

[0011] (3) Based on the embedded word vector representation of each checkpoint in step (2), the importance of each checkpoint is ranked in combination with the improved PageRank algorithm;

[0012] (4) Use the node deletion method to simulate the situation where the checkpoint fails for some reason. After removing the specified checkpoint, predict the future traffic flow of all checkpoints based on the historical traffic flow data of the remaining checkpoints. Use the prediction error to test the effectiveness of the importance ranking of each checkpoint in step (3).

[0013] Preferably, the data source obtained in step (1) is the vehicle passing data of each checkpoint at different timestamps, and the vehicle passing data includes the license plate of the passing vehicle, the vehicle model, the passing time, the checkpoint name, the checkpoint location, and the road number.

[0014] Preferably, the data preprocessing in step (1) includes: cleaning and eliminating duplicate data, missing data and invalid data of the data source, counting the daily traffic volume of each checkpoint, and extracting vehicle trajectories according to certain time intervals.

[0015] Preferably, step (2) specifically includes:

[0016] (21): Define the following basic concepts:

[0017] Definition of traffic network: A traffic network is a directed graph G(V, E), where V represents a node and E represents an edge; each v∈V represents a checkpoint, and each e∈E represents a connection between two nodes;

[0018] Define trajectory: The trajectory sequence of a vehicle is t = [v1, v2, ..., v n ], where v i∈ V,n is the length of the current trajectory;

[0019] Define the trajectory corpus: trajectory corpus TCorpus = {t1, t2, ..., t M}, where M represents the number of all trajectories, and the trajectory corpus is the set of all trajectories;

[0020] Define trajectory context: For a trajectory t i =[v1, v2, ..., v n]Context of length m The trajectory context is all the consecutive subtrajectories of length m in the trajectory;

[0021] Define the trajectory context corpus: the length of the trajectory context corpus is m

[0022] Define node context: For a node v i , whose context is of length m and At the same time, define v i for The central word of

[0023] Define node embedding representation: For a certain v∈V, the node embedding representation θ v =(a1, a2, ..., a k )∈R k , is the dimension of a pre-given vector;

[0024] (22) According to the definition of step (21), the Glove model in natural language processing is used to embed the word vector representation of the card slot; assuming that the co-occurrence matrix is ​​X, X i Indicates bayonet i, X ij Represents the number of times that checkpoints j and i co-occur within a context window of a specific size. Based on the distance d between the two checkpoints in the context window, a decay function, i.e., decay = 1 / d, is used to calculate the weight. An approximate relationship between word vectors and co-occurrence matrices is constructed:

[0025]

[0026] Where w i and w j is the required word vector, b i and b j They are the bias items of the two word vectors respectively;

[0027] Finally, construct the loss function:

[0028]

[0029] Where: |V| is the size of the dictionary, the computational complexity of J is linearly related to the number of non-zero elements of the co-occurrence matrix X, and f(X ij ) is the weight item.

[0030] Preferably, step (3) specifically includes:

[0031] (31) Define the following basic concepts:

[0032] Definition of multi-layer temporal traffic network: A multi-layer temporal traffic network is a directed graph G(N, E, K), where N and E are the number of nodes and edge sets in the multi-layer temporal traffic network, respectively, and K = (K 1 , K 2 ,…,K L ) is a series of subnets K α =(N,E α )(α={1,2,…,L}); the same nodes establish inter-layer directed connections in chronological order, and the inter-layer link from layer α to layer β is defined as Where β = α + 1;

[0033] Define the bayonet adjacency matrix A: if bayonet u is adjacent to bayonet v, then Auv = 1, otherwise it is 0;

[0034] Define the bayonet node similarity matrix S: S = λ·S in +μ·S out +ν· semantic , where λ+μ+ν=1, S in is the in-degree similarity, S out is the out-degree similarity, S semantic is semantic similarity;

[0035] In a multi-layer temporal traffic network, the similarity S between a checkpoint node u and its upstream and downstream checkpoint nodes v is uv It is a linear combination of the following three parts:

[0036] In-degree similarity:

[0037]

[0038] Where, α in (u) and α in (v) are the neighbor checkpoint sets pointing to checkpoint nodes u and v respectively;

[0039] Outdegree similarity:

[0040]

[0041] Where, α out (u) and α out (v) are the neighbor checkpoint sets pointed to by checkpoint nodes u and v respectively;

[0042] Semantic similarity:

[0043]

[0044] Where θ u and θ v are the embedded word vectors of the checkpoint nodes u and v respectively;

[0045] Define the biased random walk jump probability matrix P: At a certain moment, the probability that a vehicle passing through checkpoint u in network layer α moves to checkpoint v is for

[0046]

[0047] Where, represents the similarity between the checkpoints u and v in the network layer α; represents the adjacent relationship between the checkpoints u and v in the network layer α;

[0048] (32) According to the above definition, the centrality of the checkpoint nodes in the first layer of the network at the initial moment is calculated; the biased random walk is applied to the classic PageRank algorithm to obtain the biased PageRank based on the similarity of the checkpoint nodes (BSBPR) to evaluate the importance of the nodes in the first layer:

[0049]

[0050] (33) Recursively calculate the BSBPR of nodes in layers α = 2, 3, ..., L; define the multi-layer temporal biased PageRank of node u in layer α (α ≥ 2) as:

[0051]

[0052]

[0053] Where, BSBPR(u) α-1 and They represent the power of the centrality of the same checkpoint node in the upper layer and the similarity between adjacent checkpoint nodes in the layer, respectively. The exponents a and b are deviation parameters and a>0, b>0.

[0054] Preferably, step (4) specifically includes:

[0055] (41) Remove the designated checkpoints and fill the corresponding checkpoint traffic with zeros to simulate the special situation where the checkpoint fails for some reason. The total error of the traffic flow prediction of each checkpoint in the entire road network on the next day is used as the main indicator to verify the rationality and effectiveness of the algorithm in step (3);

[0056] (42) A traffic prediction model based on the machine learning method Xgboost is established, with the traffic at the checkpoints without faults in the road network on the previous day as input and the traffic at all checkpoints in the road network on the next day as output, and prediction is performed according to the method described in (41).

[0057] The beneficial effects of the present invention are:

[0058] The present invention aims to sort key road points, taking into account both the physical topology of the traffic network and traffic demand and actual travel trajectories, and objectively derives the importance of different checkpoints from multiple perspectives. To address the shortcomings of traditional sorting methods, a word vector embedding model is introduced to measure the similarity of checkpoint entry and exit degrees while mining the semantic relevance between checkpoints. Furthermore, a multi-layer temporal network is established based on the characteristics of the traffic network. The PageRank algorithm based on these improvements can well take into account the influence of history. Finally, a machine learning method, Xgboost, is established to predict traffic flow and verify the effectiveness of the sorting algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0060] Figure 1 Provide an overall flow chart of the preferred embodiment patent for the present invention;

[0061] Figure 2 This is a distribution map of all checkpoints in the cities selected in the embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0063] like Figures 1 to 2 As shown, this embodiment provides a method for ranking the importance of traffic checkpoints based on embedded word vectors and PageRank algorithm.

[0064] like Figure 1 The following is the overall flow chart of this embodiment, which includes four modules: data acquisition and preprocessing module, card slot word vector embedding module, PageRank algorithm improvement module, and traffic prediction model algorithm validity verification module. The detailed implementation process of this embodiment is as follows, including four steps:

[0065] S1: Obtaining data sources and preprocessing data: Obtain data sources from traffic management departments and related companies. The data sources include the user's vehicle information, including but not limited to license plates, vehicle models, and vehicle passing information at various checkpoints. Preprocessing is performed, specifically including:

[0066] S11: Data acquisition: Based on the original data provided by the management department and related companies, the vehicle passing data of each checkpoint at different timestamps is acquired, including the license plate, vehicle model, passing time, checkpoint name, checkpoint location, and road number;

[0067] S12: Preprocessing raw data: Clean and remove duplicate, missing, and invalid data from the data source, count the daily traffic volume at each checkpoint, and extract vehicle trajectories based on a certain time interval;

[0068] S2: Checkpoint word embedding: First, we extract the set of traffic checkpoints used to represent locations in the user's trajectory. Then, we introduce the Glove model from natural language processing and obtain the word embedding vector representation of each checkpoint by training vehicle trajectory data. Specifically, we include:

[0069] S21: Define the following basic concepts:

[0070] Definition of a traffic network: A traffic network is a directed graph G(V, E), where V represents a node and E represents an edge. Each v∈V represents a checkpoint, and each e∈E represents a connection between two nodes.

[0071] Define trajectory: The trajectory sequence of a vehicle is t = [v1, v2, ..., v n ], where v i∈ V,n is the length of the current trajectory;

[0072] Define the trajectory corpus: trajectory corpus TCorpus = {t1, t2, ..., t M}, where M represents the number of all trajectories, and the trajectory corpus is the set of all trajectories;

[0073] Define trajectory context: For a trajectory t i =[v1, v2, ..., v n ]Context of length m The trajectory context is all the consecutive subtrajectories of length m in the trajectory;

[0074] Define the trajectory context corpus: the length of the trajectory context corpus is m

[0075] Define node context: For a node v i , whose context is of length m and At the same time, define v i for The central word of

[0076] Define node embedding representation: For a certain v∈V, the node embedding representation θ v =(a1, a2, ..., a k )∈R k , is the dimension of a pre-given vector;

[0077] S22: Use the Glove model in natural language processing to embed word vector representation of the card slot; assuming the co-occurrence matrix is ​​X, X i Indicates bayonet i, x ij represents the number of times that checkpoints j and i co-occur within a context window of a specific size. Based on the distance d between the two checkpoints in the context window, a decreasing weighting function (decay = 1 / d) is used to calculate the weight. In other words, the farther the distance between the two words, the smaller the weight of the total count. Then, an approximate relationship between the word vector and the co-occurrence matrix is ​​constructed:

[0078]

[0079] Where w i and w j is the required word vector, b i and b j They are the bias items of the two word vectors respectively;

[0080] Finally, construct the loss function:

[0081]

[0082] Where: |V| is the size of the dictionary, the computational complexity of J is linearly related to the number of non-zero elements of the co-occurrence matrix X, and f(X ij ) is a weight item and must meet the following conditions:

[0083] (1) f(0) = 0, because It is limited;

[0084] (2) the proportion of less frequent co-occurrences is smaller;

[0085] (3) For co-occurrences that occur frequently, f(x) cannot be too large;

[0086] A weight function form with good experimental effect is:

[0087]

[0088] S3: Improve the PageRank algorithm steps, including:

[0089] S31: Define the following concepts:

[0090] Definition of multi-layer temporal traffic network: A multi-layer temporal traffic network is a directed graph G(N, E, K), where N and E are the number of nodes and edge sets in the multi-layer temporal traffic network, respectively, and K = (K 1 , K 2 ,…,K L ) is a series of subnets K α =(N,E α )(α={1,2,…,L}); the same nodes establish inter-layer directed connections in chronological order, and the inter-layer link from layer α to layer β is defined as Where β = α + 1;

[0091] Define the bayonet adjacency matrix A: If bayonet u is adjacent to bayonet v, then A uv =1, otherwise 0;

[0092] The similarity index between nodes is introduced as a sign to measure the local importance of nodes, and the definition of the node similarity matrix is ​​given as follows:

[0093] Define the checkpoint node similarity matrix S: In a multi-layer temporal traffic network, the similarity S between the checkpoint node u and its upstream and downstream checkpoint nodes v is: uv It is a linear combination of the following three parts:

[0094] (1) In-degree similarity:

[0095]

[0096] Where, α in (u) and α in (v) are the neighbor checkpoint sets pointing to checkpoint nodes u and v respectively.

[0097] (2) Out-degree similarity:

[0098]

[0099] Where, α out (u) and α out (v) are the neighbor checkpoint sets pointed to by checkpoint nodes u and v respectively.

[0100] (3) Semantic similarity:

[0101]

[0102] Where θ u and θ v are the embedded word vectors of the checkpoint nodes u and v respectively;

[0103] The final similarity matrix is ​​S = λ·S in +μ·S out +v·S swmantic , where λ+μ+μ=1;

[0104] Define the biased random walk jump probability matrix P: At a certain moment, the probability that a vehicle passing through checkpoint u in network layer α moves to checkpoint v is for:

[0105]

[0106] Where, represents the similarity between the checkpoints u and v in the network layer α; represents the adjacent relationship between the checkpoints u and v in the network layer α;

[0107] S32: Calculate the centrality of the checkpoint nodes in the first layer of the network at the initial moment; for this purpose, apply biased random walk to the classic PageRank algorithm to obtain biased PageRank based on checkpoint node similarity (BSBPR) to evaluate the importance of nodes in the first layer:

[0108]

[0109] S33: Recursively calculate the BSBPR of nodes in layers α = 2, 3, ..., L; combined with the ordered unidirectional connections between layers, the centrality of a node u in layer α is related to the similarity of its neighboring nodes and is also affected by the centrality of the same node in layer α-1, BPSBPR(u) α-1 The influence of ; that is, the multi-layer temporal biased PageRank of node u in layer α includes the relevant network topology information of the previous layer; therefore, the multi-layer temporal biased PageRank of node u in layer α (α≥2) is defined as:

[0110]

[0111]

[0112] Where, BSBPR(u) α-1 and They represent the power of the centrality of the same checkpoint node in the upper layer and the similarity between adjacent checkpoint nodes in the layer, respectively. The exponents a and b are deviation parameters used to adjust the degree of dependence of the node on the intra-layer and inter-layer factors, and both exponents a and b take values ​​greater than zero, that is, a>0, b>0;

[0113] S4: Traffic flow prediction step to verify the effectiveness of the algorithm. The node deletion method is used to simulate the situation where a checkpoint fails for some reason. After removing the specified checkpoint, the traffic flow of all checkpoints in the future is predicted based on the historical traffic flow data of the remaining checkpoints. The effectiveness of the checkpoint sorting algorithm is verified by the prediction error. Specifically, the following steps are performed:

[0114] S41: Remove the designated checkpoints and fill the corresponding checkpoint traffic with zeros to simulate the special situation where the checkpoint fails for some reason, and use the total error of the traffic prediction of each checkpoint in the entire road network on the next day as the main indicator to verify the rationality and effectiveness of the algorithm we proposed. According to common sense, when there are a large number of faulty checkpoints in the road traffic network on a certain day, since we cannot obtain the traffic flow information of the corresponding checkpoints on that day, the error of the traffic flow predicted for each checkpoint in the entire road network on the next day should be larger. When there are a small number of faulty checkpoints in the road traffic network on a certain day, although we still cannot obtain the traffic flow information of all checkpoints on that day, since most of the checkpoints are normal and we have considered the correlation between the checkpoints in the newly proposed algorithm, the error of the traffic flow predicted for each checkpoint in the entire road network on the next day should be smaller.

[0115] S42: Establish a traffic prediction model based on the machine learning method Xgboost, use the traffic at the checkpoints without faults in the road network on the previous day as input, and the traffic at all checkpoints in the road network on the next day as output, and perform prediction according to the method described in S41.

[0116] This embodiment fully leverages the physical topology of the transportation network, traffic demand, and actual travel trajectories to more objectively determine the importance of different checkpoints from multiple perspectives. This embodiment trains a word vector model based on trajectory data to obtain embedded word vector representations of checkpoints in the road network. Furthermore, in real life, downstream checkpoints are generally highly correlated with upstream checkpoints—that is, a vehicle passing through an upstream checkpoint is highly likely to pass through its downstream checkpoint. Therefore, the degree correlation and semantic relevance between checkpoints should be fully considered. Furthermore, this embodiment establishes a machine learning-based traffic flow prediction model based on the node deletion method to verify the validity and rationality of the ranking results.

[0117] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A traffic checkpoint importance ranking method based on word embedding vectors and PageRank algorithm, characterized by: The following steps are involved: (1) Obtain the data source including the user's vehicle information and the collection of vehicle passing information at each checkpoint, and pre-process the data from the data source to extract the user's vehicle trajectory; (2) Extract the set of traffic checkpoints used to represent the location in the user's vehicle trajectory in step (1); introduce the Glove model in natural language processing, and obtain the embedded word vector representation of each checkpoint by training the vehicle trajectory data; (3) Based on the embedded word vector representation of each checkpoint in step (2), the importance of each checkpoint is ranked in combination with the improved PageRank algorithm; dynamic modeling of the multi-layer temporal traffic network is carried out, specifically including: (31) The following basic concepts are defined: Define a multi-layer temporal traffic network: A multi-layer temporal traffic network is a directed graph , where N and E are the number of nodes and edge sets in the multi-layer temporal traffic network, is a series of subnets The same nodes establish directed connections between layers in chronological order, defining Layer Direction The inter-layer links are ,in ; Define the card port adjacency matrix A: If the card port With bayonet Adjacent, then , otherwise 0; Define the bayonet node similarity matrix S: ,in , is the in-degree similarity, is the out-degree similarity, is semantic similarity; In a multi-layer time-series traffic network, checkpoint nodes and its upstream and downstream bayonet nodes Similarity It is a linear combination of the following three parts: In-degree similarity: Where, and Pointing to the card port node and The neighbor card set; Outdegree similarity: Where, and Bayonet nodes and the set of neighboring bayonets pointed to by v; Semantic similarity: Where, and Bayonet nodes and the embedded word vector of v; Define the biased random walk jump probability matrix P: At a certain moment, the network layer Through the bayonet The probability that a vehicle moves toward checkpoint v for Where, Represents the network layer The similarity between the middle bayonet u and v; Represents the network layer The adjacent relationship between the middle bayonet u and v; (32) According to the above definition, the centrality of the checkpoint nodes in the first layer of the network at the initial moment is calculated; the biased random walk is applied to the classic PageRank algorithm to obtain the biased PageRank based on the similarity of the checkpoint nodes (BSBPR) to evaluate the importance of the nodes in the first layer: (33) Recursively in layers Compute nodes in ;Define layer midpoint The multi-layer temporal biased PageRank is: Where, and Respectively represent the power of the centrality of the same checkpoint node in the upper layer and the similarity between adjacent checkpoint nodes in the layer, where the exponent and is the bias parameter and , ; (4) Use the node deletion method to simulate the situation where the checkpoint fails for some reason. After removing the specified checkpoint, predict the traffic flow of all checkpoints in the future based on the historical traffic flow data of the remaining checkpoints. Use the Xgboost model and the correlation analysis of the historical traffic flow data to test the prediction error. Step (3) The effectiveness of the importance ranking of each checkpoint includes: (41) Remove the designated checkpoints and fill the corresponding checkpoint traffic with zeros to simulate the special situation where the checkpoint fails for some reason. The total error of the traffic flow prediction of each checkpoint in the entire road network on the next day is used as the main indicator to verify the rationality and effectiveness of the algorithm in step (3); (42) A traffic prediction model based on the machine learning method Xgboost is established, with the traffic at the checkpoints without faults on the road network on the previous day as input and the traffic at all checkpoints on the road network on the next day as output, and prediction is performed according to the method described in (41).

2. The method for ranking traffic checkpoint importance based on word embedding vectors and PageRank algorithm as claimed in claim 1, characterized in that: The data source obtained in step (1) is the vehicle passing data of each checkpoint at different time stamps, and the vehicle passing data includes the license plate of the vehicle, the vehicle model, the time of passing, the checkpoint name, the checkpoint location, and the road number.

3. The method for ranking traffic checkpoint importance based on word embedding vectors and PageRank algorithm as claimed in claim 1, characterized in that: The data preprocessing in step (1) includes: cleaning and eliminating duplicate data, missing data, and invalid data from the data source, counting the daily traffic volume at each checkpoint, and extracting vehicle trajectories according to certain time intervals.

4. The method for ranking traffic checkpoint importance based on word embedding vectors and PageRank algorithm as claimed in claim 1, characterized in that: Step (2) specifically includes: (21): Define the following basic concepts: Definition of transportation network: transportation network is a directed graph ,in represents a node, E represents an edge; each Indicates a bayonet, each Represents a connection between two nodes; Define trajectory: The trajectory sequence of a vehicle is ,in , is the length of the current trajectory; Defining the trajectory corpus: Trajectory Corpus ,in Represents the number of all trajectories, and the trajectory corpus is the set of all trajectories; Define track context: For a track Length is Context , the trajectory context is the length of all consecutive sub-trajectory of ; Define the trajectory context corpus: length is Trajectory Context Corpus ; Define node context: for a node , whose length is Context ,and At the same time, define for The central word of Define node embedding representation: for a , the embedded representation of the node , , is the dimension of a pre-given vector; (22) According to the definition of step (21), the Glove model in natural language processing is used to embed the word vector representation of the card slot; assuming that the co-occurrence matrix is , Indicates the bayonet , Indicates the bayonet and bayonet The number of co-occurrences within a context window of a specific size; based on the distance between the two bayonet points in the context window , using the decay function Used to calculate weights; build an approximate relationship between word vectors and co-occurrence matrices: Where, and is the required word vector, and They are the bias items of the two word vectors respectively; Finally, construct the loss function: Where: |V| is the size of the dictionary, The computational complexity and co-occurrence matrix of The number of non-zero elements of is linear, is the weight item.

Images