Vent travel big data visualization method and system

Abstract

The invention belongs to the field of visualization, and provides a text travel big data visualization method and system, and the method comprises the steps: collecting text travel related high-dimensional data; carrying out dimension reduction processing on the text travel related high-dimensional data, mapping the high-dimensional data to a low-dimensional space to obtain low-dimensional data, and in the dimension reduction process, keeping the distribution change of the data not to exceed a first threshold value by using a first algorithm; data points in the low-dimensional data serve as nodes in a network graph, each node represents a data object, edges between the nodes are determined according to the relevance between the data, and the weights of the edges are dynamically adjusted through a graph attention network based on the weights of the relevance edges between the data objects; determining the distance between the nodes based on the weight of the edge, and adjusting the relative position between the nodes by using a second algorithm to obtain a first node layout; and displaying the first node layout by using an interactive visualization tool.

Description

Technical Field

[0001] The present invention belongs to the field of visualization, and specifically relates to a method and system for visualizing cultural and tourism big data. Background Art

[0002] With the rapid development of tourism, the cultural and tourism industry has gradually formed a massive and multi-dimensional data system. These data cover many aspects such as tourist behavior data, scenic spot operation data, social media comment data, etc., involving a wide variety of information with high temporal and spatial dependence and correlation. The effective integration and visualization of these data are of great significance to improving the management efficiency, service quality and decision-making ability of the cultural and tourism industry. However, in the face of huge and complex high-dimensional data, how to perform efficient data processing, dimensionality reduction, visualization and dynamic analysis has become a core technical problem in the current field of cultural and tourism big data applications.

[0003] In the existing technology, most data visualization methods focus on the display of two-dimensional or three-dimensional space, usually relying on simple dimensionality reduction technology or static graphic display. Although these methods can display the structure and relationship of data to a certain extent, they are often unable to effectively process large-scale high-dimensional cultural and tourism data, and cannot reflect the deeper relationship between data. Although existing dimensionality reduction technologies (such as PCA, t-SNE, etc.) can map high-dimensional data to low-dimensional space, there are still problems such as insufficient information retention and high computational complexity.

[0004] In addition, visualization technology based on graph models is widely used in social networks, transportation networks and other fields. However, in the field of culture and tourism, especially in the context of multi-dimensional and diversified cultural and tourism data, how to design an effective graph structure and how to dynamically adjust the weights of nodes and edges in the graph is still a technical problem that needs to be further solved. Summary of the invention

[0005] In order to solve the problems in the prior art, the present invention provides a method for visualizing cultural and tourism big data, which comprises the following steps:

[0006] Collecting high-dimensional data related to culture and tourism, including tourist behavior data, scenic spot operation data, and social media comment data;

[0007] Performing dimensionality reduction processing on the cultural tourism related high-dimensional data, mapping the high-dimensional data to a low-dimensional space to obtain low-dimensional data, and in the dimensionality reduction process, using a first algorithm to keep the distribution change of the data from exceeding a first threshold;

[0008] The data points in the low-dimensional data are used as nodes in a network graph, each node represents a data object, the edges between the nodes are determined according to the correlation between the data, and the weights of the edges are dynamically adjusted using a graph attention network based on the weights of the edges of the correlation between the data objects;

[0009] Determine the distance between the nodes based on the weight of the edge, and use a second algorithm to adjust the relative positions between the nodes to obtain a first node layout;

[0010] An interactive visualization tool is used to display the first node layout, using color, node size, and edge width attributes to distinguish different attributes of nodes and relationship strengths.

[0011] Another aspect of the present invention further provides a cultural tourism big data visualization system, the system comprising the following modules:

[0012] A collection module is used to collect high-dimensional data related to culture and tourism, including tourist behavior data, scenic spot operation data, and social media comment data;

[0013] A dimension reduction module is used to perform dimension reduction processing on the cultural tourism related high-dimensional data, map the high-dimensional data to a low-dimensional space, and obtain low-dimensional data. During the dimension reduction process, a first algorithm is used to keep the distribution change of the data from exceeding a first threshold;

[0014] A first updating module is used to use the data points in the low-dimensional data as nodes in a network graph, each node represents a data object, determine the edges between the nodes according to the correlation between the data, and dynamically adjust the weights of the edges based on the weights of the correlation edges between the data objects using a graph attention network;

[0015] A second updating module, configured to determine the distance between the nodes based on the weight of the edge, and use a second algorithm to adjust the relative positions between the nodes to obtain a first node layout;

[0016] The display module is used to use an interactive visualization tool to display the first node layout, and use color, node size, and edge width attributes to distinguish different attributes of nodes and relationship strengths.

[0017] The cultural tourism big data visualization method of this invention has the following beneficial effects:

[0018] The present invention can effectively map cultural and tourism related high-dimensional data to low-dimensional space, which not only retains the global structure of the data, but also maintains the accuracy of local features. This process ensures that the difference between the distribution of the data after dimensionality reduction and the distribution of the original data is controlled within a reasonable range by optimizing the stress function, effectively solving the problem of information loss in the process of dimensionality reduction of high-dimensional data.

[0019] By using the data points mapped from low-dimensional data as nodes in the network graph and dynamically adjusting the weights of the edges between nodes in combination with the graph attention network, the correlation between data can be accurately reflected. In particular, based on the actual situation of cultural and tourism data, the relationship between nodes is no longer static, but is adjusted in real time as the data changes, thereby improving the flexibility and expressiveness of the graph model.

[0020] The interactive visualization tool used in the present invention can dynamically display the first node layout, and use attributes such as color, node size, and edge width to distinguish the attributes and relationship strength of different nodes. This allows decision makers to intuitively understand the structural characteristics of cultural and tourism data, quickly identify potential hot spots, tourist behavior patterns, and scenic spot operation conditions, thereby providing a scientific basis for optimizing operations and improving service quality.

[0021] The cultural tourism big data visualization method of the present invention is not only applicable to scenic spot operation and tourist behavior analysis, but can also be widely used in other cultural tourism fields, such as travel route optimization, tourist preference prediction, market analysis, etc. Its modeling method based on graph structure and dimensionality reduction technology make the method more flexible and extensible, and can adapt to different types of cultural tourism data analysis needs.

[0022] Through the implementation of the above-mentioned technical solutions, decision makers in the cultural and tourism industry will be able to more accurately identify trends and potential problems in the data, thereby improving the intelligence level of cultural and tourism operations, enhancing the tourist experience, and promoting the sustainable development and innovation of the cultural and tourism industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 work.

[0024] Figure 1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0025] Below, the invention is preferably described in conjunction with the accompanying drawings and specific implementation methods.

[0026] This embodiment solves the above problem through the following steps:

[0027] In one embodiment, reference Figure 1The present invention provides a method for visualizing cultural tourism big data, which aims to comprehensively process and display multi-source and multi-dimensional cultural tourism data, reveal the intrinsic connections and dynamic characteristics between the data, and thus provide scientific support for the operation optimization, tourist behavior analysis and market decision-making of the cultural tourism industry. This method realizes a panoramic presentation of cultural tourism data through steps such as high-dimensional data dimensionality reduction, node graph construction, and interactive chart display. Specifically, the method includes the following steps:

[0028] Collect high-dimensional data related to culture and tourism, including tourist behavior data, scenic spot operation data, and social media comment data.

[0029] Collect high-dimensional data related to the cultural and tourism industry, including but not limited to the following important dimensions:

[0030] Tourist behavior data includes the various activities of tourists in and outside the scenic area, length of stay, frequency of visits, consumption records, etc. For example, the movement path of tourists in the scenic area can be tracked in real time through positioning devices (such as GPS, Wi-Fi positioning, etc.), recording the movement trajectory of tourists between various scenic spots and the length of stay at specific scenic spots. In addition, tourists' consumption behavior data, such as spending records on dining, souvenir purchases, etc., can also provide insights into tourists' preferences and help scenic spots analyze tourists' consumption patterns and trends.

[0031] Scenic area operation data covers the daily operation status of scenic areas, including the number of tourists received, scenic area opening hours, facility usage, equipment failure and maintenance records, tourist satisfaction feedback, etc. For example, the number of tourists received by major scenic spots in the scenic area and its changing trends can help managers predict tourist flow and optimize resource allocation. Equipment maintenance records can reflect the operating status of equipment in the scenic area and promptly identify potential operational risks.

[0032] Social media review data comes from comments, feedback, ratings, photos and video content posted by tourists on social platforms (such as Weibo, Douyin, TripAdvisor, etc.). These data can not only be used for sentiment analysis through natural language processing (NLP) technology to obtain tourists' overall satisfaction and emotional tendencies towards scenic spots, but also to extract hot topics, attractions recommended by tourists, and their main concerns. For example, by analyzing the keywords in the comments, tourists' favorite attractions or services can be identified to help scenic spots optimize the travel experience.

[0033] Through the collection and integration of these high-dimensional data, this method can comprehensively reveal the behavior patterns of tourists, the operational efficiency of scenic spots, and tourists' feedback on scenic spots, thereby providing accurate decision-making basis for scenic spot managers and improving the overall tourism service level and tourist satisfaction.

[0034] The cultural and tourism related high-dimensional data is subjected to dimensionality reduction processing, and the high-dimensional data is mapped to a low-dimensional space to obtain low-dimensional data. During the dimensionality reduction process, a first algorithm is used to keep the distribution change of the data within a first threshold.

[0035] Cultural and tourism-related high-dimensional data, such as tourist behavior data, scenic spot operation data, and social media comment data, usually contain a large number of features, which are often difficult to directly and effectively analyze. In order to facilitate analysis and improve computational efficiency, the present invention uses dimensionality reduction technology to map high-dimensional data to low-dimensional space while retaining the global and local structural information of the data. The dimensionality reduction process uses a first algorithm to ensure that the distribution change of the data does not exceed a first threshold, thereby effectively reducing the dimension while maximizing the characteristics of the original data.

[0036] To ensure that the distribution of the data does not change more than the first threshold, first, for each pair of data points x i and x j (where i, j = 1, 2, ..., N, N is the total number of data points), calculate their distance δ in high-dimensional space ij

[0037]

[0038] in, Represents the value of the i-th data point in the k-th dimension, and d is the number of dimensions of the data.

[0039] Use the t-SNE (t-distributed stochastic neighbor embedding) algorithm to randomly generate an initial position y in the low-dimensional space for each high-dimensional data point i , and set the target dimension d' of the low-dimensional space. Preferably, d' is 3 dimensions. In the low-dimensional space, calculate each pair of low-dimensional data points y i and j The distance between ij

[0040]

[0041] in Represents the coordinate of the i-th data point in the k-th dimension in the low-dimensional space.

[0042] Define the difference function S to quantify the distance d between low-dimensional data points ij The distance δ between the high-dimensional data points ij The degree of difference between

[0043]

[0044] The smaller the difference value is, the closer the low-dimensional data mapping is to the structure of the high-dimensional data.

[0045] The gradient descent algorithm is used to adjust the coordinates of the low-dimensional data points so that the distance between the low-dimensional data points is as close as possible to the distance in the high-dimensional space. The update formula is as follows:

[0046]

[0047] in represents the coordinates of the i-th data point in the k-th dimension, η is the learning rate, is the difference function about the low-dimensional data point position y i Gradient

[0048]

[0049] This formula guides the gradient descent process by calculating the update direction for each low-dimensional data point position.

[0050] The dimensionality reduction process is performed iteratively until the difference function S converges to the preset threshold ε1, that is, the condition S≤ε1 is satisfied, or the maximum number of iterations Tmax is reached.

[0051] In the dimensionality reduction process, by controlling the change of the difference function S, the change of the data distribution is ensured not to exceed the first threshold ε1, thereby effectively retaining the structural characteristics of the high-dimensional data. In the optimization process, when the difference function is less than or equal to the first threshold, the dimensionality reduction process is completed and a low-dimensional data set is generated.

[0052] Through the above steps, the present invention can realize efficient dimensionality reduction processing of cultural and tourism big data, reduce computational complexity, and at the same time maintain the intrinsic relationship and structural characteristics of the data, providing a reliable foundation for subsequent data analysis and visualization.

[0053] The data points in the low-dimensional data are used as nodes in a network graph, each node represents a data object, and the edges between the nodes are determined according to the correlation between the data. Based on the weight of the edge of the correlation between the data objects, the weight of the edge is dynamically adjusted using a graph attention network.

[0054] Each data point y in the low-dimensional data after dimensionality reduction processing in the cultural and tourism big data i Mapped to a node v in the network graph i , where i = 1, 2, ..., N, and N is the total number of data points. Each node v i Represents a specific cultural tourism data object, such as the behavior data of a tourist, the operating indicators of a scenic spot, or the comment data of a social media. The attributes of the node include but are not limited to the basic information of tourists, the facilities of the scenic spot, the sentiment score of social media comments, etc.

[0055] The edge between nodes ij Represents node v i With node vj Specifically, the edge generation rules are based on the similarities or relationships between different data objects in the cultural and tourism big data. For example, the correlation between tourist behavior data and scenic spot operation data can be measured by common points of interest, scenic spot visit frequency and other indicators. The edge construction formula is as follows:

[0056] e ij =f(y i ,y j ), where f(y i ,y j ) is the measurement node v i and node v j The similarity function between the two nodes can use common similarity metrics including Euclidean distance, cosine similarity, etc. Specifically, if the tourist behavior data and scenic spot operation data represented by the two nodes are highly correlated, an edge is established, and the correlation degree of the edge is high.

[0057] On the basis of constructing the edge, a weight w is assigned to each edge according to the correlation between nodes. ij , which reflects the strength of association between data objects. For example, if the length of stay and the amount of consumption of tourists represented by two nodes in a scenic spot are similar, the edge e can be given ij Assign a larger weight. The weight calculation formula is as follows:

[0058]

[0059] Among them, σ is a regularization parameter that controls the degree of smoothing of the weights.

[0060] In order to dynamically adjust the edge weights, the Graph Attention Network (GAT) is used to update the edge weights. The Graph Attention Network introduces an attention mechanism to dynamically assign different weights to each edge in each layer to emphasize more relevant neighboring nodes. Specifically, during the propagation of node features, the Graph Attention Network calculates the attention coefficient α between nodes. ij , which reflects the node v i For node v j The extent of the impact

[0061]

[0062] Where a is the learned weight vector, T represents the transpose, and N(i) represents the node v i The neighbor set of , || represents the vector connection operation, and LeakyReLU is the activation function.

[0063] According to the attention coefficient α ij , node v i Update its feature y i, the new node feature is obtained by weighted summation of the features of neighboring nodes:

[0064]

[0065] where y i′ is a new node feature, σ is an activation function, and ReLU or Sigmoid can be used in the present invention. In this way, the graph attention network can dynamically adjust the dependency between nodes according to the weight of each edge.

[0066] The edge weights adjusted by the graph attention network can more accurately reflect the potential correlation between cultural and tourism data objects. For example, between social media comment data and scenic spot operation data, some comments may significantly affect tourists' behavior patterns. The graph attention network can capture these influence relationships by adjusting edge weights, thereby providing valuable basis for related analysis of the cultural and tourism industry (such as tourist preference prediction, scenic spot optimization, etc.).

[0067] Through the above steps, the present invention can construct a multi-level, dynamically adjusted network graph based on cultural and tourism big data, and optimize and adjust the edge weights through the graph attention network, thereby realizing accurate modeling and efficient analysis of cultural and tourism data, improving the effects of data mining and predictive analysis, and providing strong support for the development of the cultural and tourism industry.

[0068] The distance between the nodes is determined based on the weight of the edge, and the relative positions between the nodes are adjusted using a second algorithm to obtain a first node layout.

[0069] In the network diagram built based on cultural and tourism big data, node v i and node v j The edge weight w between ij It reflects the strength of the association between them. In order to effectively display the relationship between cultural and tourism data objects in graph visualization, the edge weight w ij is used to determine the node v i and node v j Specifically, the edge weight is related to the distance d between nodes. ij There is an inverse relationship, that is, the stronger the association, the smaller the distance between nodes. It can be mapped by the following formula:

[0070]

[0071] Here, ∈ is a small constant used to avoid division by zero errors. This formula shows that the larger the edge weight (i.e., the higher the node association), the smaller the physical distance between them. For tourist behavior data, scenic spot operation data, and social media comment data in cultural and tourism big data, if the tourist behavior patterns represented by two nodes are similar or the tourist activities in the scenic spot are frequently related, the distance between the two nodes in the graph will be small.

[0072] After constructing the network graph, the initial node positions (first node layout) can be determined by the edge weight mapping described above. The initial node layout will be initialized based on the data relationship reflected by the edge weights through a simple physical model, such as the spring-charge model. In this layout, the distance between nodes is determined by the edge weights w ij Directly determined by this, the goal of the initial layout is to cluster related data objects together as much as possible in order to better reflect the structural characteristics of cultural and tourism data.

[0073] In order to optimize the graph layout, make the relative positions between nodes more reasonable, and visually reflect the potential relationship between data objects, the second algorithm is used to adjust the initial node layout. The goal of the second algorithm is to minimize the energy function between nodes, so that the position of the nodes is more consistent with the actual correlation strength between data.

[0074] Specifically, the second algorithm determines the positions of nodes by simulating the interaction forces between them. In the application of cultural and tourism big data, the relationship between nodes is converted into forces, and the nodes interact with each other through attraction and repulsion, thus eventually forming a reasonable layout. The core idea of ​​the algorithm is to make nodes with larger edge weights close to each other, while nodes with smaller edge weights stay away from each other.

[0075] If the edge weight w between two nodes ij The larger the value, the stronger the attraction between the two nodes, which means they should be closer in the graph. The magnitude of the attraction can be expressed as:

[0076] F attr,ij =k1·w ij ·log(d ij )

[0077] Where k1 is a constant that controls the strength of attraction, log(d ij ) is a logarithmic function that balances the rate at which attraction changes with distance.

[0078] In order to avoid nodes being too dense, a repulsive force is introduced. The distance d between the repulsive force and the node ij The repulsive force is inversely proportional to the distance between nodes. When the nodes are too close, the repulsive force increases, pushing the nodes away from each other. The magnitude of the repulsive force can be expressed as:

[0079]

[0080] Among them, k2 is a constant that controls the strength of the repulsive force.

[0081] Total force between nodes F ij is the combination of attractive and repulsive forces, expressed as:

[0082] F ij =F attr,ij +F repel,ij

[0083] By solving the total force between all nodes, the movement direction and speed of each node can be obtained, thereby updating the position of the node.

[0084] According to the total force on each node, update the node coordinates (x i ,y i ), so that it moves in the direction of the force. The update formula of the node position is as follows:

[0085]

[0086] in, is the new coordinate, Δt is the time step, m i is the mass of the node, which is taken as a constant in the present invention.

[0087] Through the above steps, the present invention can effectively construct and optimize the network diagram based on cultural and tourism big data, and dynamically adjust the node layout, keep the distance and weight association of each node in the view, and the distance between nodes is neither too large nor too small.

[0088] An interactive visualization tool is used to display the first node layout, using color, node size, and edge width attributes to distinguish different attributes of nodes and relationship strengths.

[0089] In order to more intuitively display the node layout and internal relationships in the cultural and tourism big data network diagram, the present invention uses an interactive visualization tool, combined with graphical interface elements such as color, node size, edge width and other attributes, to dynamically present the different attributes of each node and edge in the network diagram and the strength of the relationship between them. Specifically, the interactive visualization tool can not only display the structural information of the nodes and edges of the network diagram, but also allow users to interact with the graphics and flexibly adjust the graphics display mode, thereby facilitating analysis and decision-making.

[0090] Different data types in cultural and tourism big data are distinguished by the color of the nodes. For example, the color of the nodes can be used to represent different cultural and tourism entity types or data sources.

[0091] It can be represented by green, indicating that the node represents a data object based on tourist behavior analysis, such as the length of stay of tourists, frequency of visits, etc.

[0092] The blue color represents data related to scenic spot operations, such as scenic spot visitor flow, facility usage, etc.

[0093] Orange is used to represent social media comment information, such as sentiment analysis results of tourist comments, keyword phrases of comments, etc.

[0094] Different colors allow each node in the diagram to quickly distinguish the type of data or data source it represents, helping users quickly identify key information.

[0095] The size of a node can be used to represent the relative importance or influence of a data object. Specifically, the size of a node can be proportional to the value of the node's attribute:

[0096] The size of the node is determined based on the number of tourists or the frequency of tourists’ visits. The larger the size of the node, the more frequent the tourist activities in the area, and vice versa.

[0097] The node size is adjusted based on important indicators such as the scenic spot's operating efficiency and revenue. For example, the node size of a well-operated scenic spot is larger, while the node size of a poorly operated scenic spot is smaller.

[0098] The size of the node can be set according to the number or sentiment of social media comments. Scenic spots with a large number of positive comments can be set as larger nodes, while scenic spots with more negative comments can be set as smaller nodes.

[0099] Through the changes in node size, users can quickly identify which nodes in the entire network diagram represent cultural and tourism data with higher importance or influence.

[0100] The width of an edge can reflect the strength or closeness of the relationship between nodes, and is particularly suitable for showing the correlation or dependency between data objects. In cultural and tourism data, the width of an edge can be adjusted according to the following criteria:

[0101] Assuming that there is a strong correlation between the tourist behavior data nodes and the scenic spot operation data nodes (for example, tourists visit a scenic spot with a high density, which leads to changes in the scenic spot's operation data), then the edge width between these nodes will be wider, indicating that the relationship between them is stronger. Conversely, if the correlation between the tourist behavior data and the scenic spot operation data is weak, the edge width will be narrower.

[0102] If there is a close connection between social media comments and scenic spot operation data (for example, negative comments on social media directly affect the operation of the scenic spot), the corresponding edge width will also increase accordingly. The change in edge width can help users intuitively identify which comment data has a greater impact on the operation of the scenic spot.

[0103] The change in edge width reflects the strength of the relationship between nodes, making the visualization of the network graph more hierarchical and able to effectively express the correlation between different data.

[0104] On the other hand, the present invention also provides a cultural tourism big data visualization system, comprising:

[0105] A collection module is used to collect high-dimensional data related to culture and tourism, including tourist behavior data, scenic spot operation data, and social media comment data;

[0106] A dimension reduction module is used to perform dimension reduction processing on the cultural tourism related high-dimensional data, map the high-dimensional data to a low-dimensional space, and obtain low-dimensional data. During the dimension reduction process, a first algorithm is used to keep the distribution change of the data from exceeding a first threshold;

[0107] A first updating module is used to use the data points in the low-dimensional data as nodes in a network graph, each node represents a data object, determine the edges between the nodes according to the correlation between the data, and dynamically adjust the weights of the edges based on the weights of the correlation edges between the data objects using a graph attention network;

[0108] A second updating module, configured to determine the distance between the nodes based on the weight of the edge, and use a second algorithm to adjust the relative positions between the nodes to obtain a first node layout;

[0109] The display module is used to use an interactive visualization tool to display the first node layout, and use color, node size, and edge width attributes to distinguish different attributes of nodes and relationship strengths.

[0110] The prior art mentioned in the above background technology section and specific embodiment section of the present invention can be used as part of the present invention to understand the meaning of some technical features or parameters.

Claims

1. A method for visualizing cultural and tourism big data, characterized in that: The method comprises the following steps: Collecting high-dimensional data related to culture and tourism, including tourist behavior data, scenic spot operation data, and social media comment data; Performing dimensionality reduction processing on the cultural tourism related high-dimensional data, mapping the high-dimensional data to a low-dimensional space to obtain low-dimensional data, and in the dimensionality reduction process, using a first algorithm to keep the distribution change of the data from exceeding a first threshold; The data points in the low-dimensional data are used as nodes in a network graph, each node represents a data object, the edges between the nodes are determined according to the correlation between the data, and the weights of the edges are dynamically adjusted using a graph attention network based on the weights of the edges of the correlation between the data objects; Determine the distance between the nodes based on the weight of the edge, and use a second algorithm to adjust the relative positions between the nodes to obtain a first node layout; An interactive visualization tool is used to display the first node layout, using color, node size, and edge width attributes to distinguish different attributes of nodes and relationship strengths.

2. The method for visualizing cultural and tourism big data according to claim 1, characterized in that: The using the first algorithm to keep the distribution change of the data not exceeding the first threshold comprises: For every pair of data points x i and x j , where i, j = 1, 2, ..., N, N is the total number of data points, and their distance δ in high-dimensional space is calculated ij in, Represents the value of the i-th data point in the k-th dimension, and d is the number of dimensions of the data; Using the t-SNE algorithm, a random initial position y is generated in the low-dimensional space for each high-dimensional data point i , and set the target dimension d' of the low-dimensional space. In the low-dimensional space, calculate each pair of low-dimensional data points y i and j The distance between ij in Represents the coordinates of the i-th data point in the k-th dimension in the low-dimensional space; Define the difference function S The gradient descent algorithm is used to update the coordinates of low-dimensional data points. The update formula is as follows: in represents the coordinates of the i-th data point in the k-th dimension, η is the learning rate, is the difference function about the low-dimensional data point position y i Gradient The dimensionality reduction process is performed iteratively until the difference function S converges to the preset threshold ε1, or the maximum number of iterations Tmax is reached.

3. The method for visualizing cultural and tourism big data according to claim 2, characterized in that: The step of dynamically adjusting the weight of the edge based on the degree of association between data objects using a graph attention network includes: For each data point y in the low-dimensional data i Mapped to a node v in the network graph i , where i = 1, 2, ..., N, and N is the total number of data points; Each node v i Represents a specific cultural and tourism data object; The edge between nodes ij Represents node v i With node v j The correlation between The edge construction formula is as follows: e ij =f(y i ,y j ), where f(y i ,y j ) is the measurement node v i and node v j A function of the similarity between According to the degree of association between nodes, a weight w is assigned to each edge ij Among them, σ is a regularization parameter that controls the smoothness of the weight; Calculate the attention coefficient α between nodes ij Where a is the learned weight vector, T represents the transpose, and N(i) represents the node v i The neighbor set of , || represents the vector connection operation, and LeakyReLU is the activation function; According to the attention coefficient α ij , node v i Update its feature y i , the new node feature is obtained by weighted summation of the features of neighboring nodes: where y i′ is the new node feature, and σ is the activation function.

4. The method for visualizing cultural and tourism big data according to claim 3 is characterized in that: Determining the distance between the nodes based on the weight of the edge includes: The edge weight and the distance between nodes d ij The mapping is done by the following formula: Here, ∈ is a constant used to avoid division by zero errors.

5. The method for visualizing cultural and tourism big data according to claim 4, characterized in that: The using the second algorithm to adjust the relative positions between the nodes to obtain the first node layout comprises: The magnitude of attraction is defined as: F attr,ij =k1·w ij ·log(d ij ) Where k1 is a constant that controls the strength of attraction, log(d ij ) is a logarithmic function that is used to balance the rate at which the attractive force changes with distance; The magnitude of the repulsive force is defined as: Among them, k2 is a constant that controls the strength of the repulsive force; Total force between nodes F ij It is the combination of attractive and repulsive forces and is expressed as: F ij =F attr,ij +F repel,ij The update formula of the node position is as follows: in, is the new coordinate, Δt is the time step, m i is the mass of the node.

6. A cultural tourism big data visualization system, characterized in that: The system includes the following modules: A collection module is used to collect high-dimensional data related to culture and tourism, including tourist behavior data, scenic spot operation data, and social media comment data; A dimension reduction module is used to perform dimension reduction processing on the cultural tourism related high-dimensional data, map the high-dimensional data to a low-dimensional space, and obtain low-dimensional data. During the dimension reduction process, a first algorithm is used to keep the distribution change of the data from exceeding a first threshold; A first updating module is used to use the data points in the low-dimensional data as nodes in a network graph, each node represents a data object, determine the edges between the nodes according to the correlation between the data, and dynamically adjust the weights of the edges based on the weights of the correlation edges between the data objects using a graph attention network; A second updating module, configured to determine the distance between the nodes based on the weight of the edge, and use a second algorithm to adjust the relative positions between the nodes to obtain a first node layout; The display module is used to use an interactive visualization tool to display the first node layout, and use color, node size, and edge width attributes to distinguish different attributes of nodes and relationship strengths.

7. The cultural and tourism big data visualization system according to claim 6, characterized in that: The using the first algorithm to keep the distribution change of the data not exceeding the first threshold comprises: For every pair of data points x i and x j , where i, j = 1, 2, ..., N, N is the total number of data points, and their distance δ in high-dimensional space is calculated ij in, Represents the value of the i-th data point in the k-th dimension, and d is the number of dimensions of the data; Using the t-SNE algorithm, a random initial position y is generated in the low-dimensional space for each high-dimensional data point i , and set the target dimension d' of the low-dimensional space. In the low-dimensional space, calculate each pair of low-dimensional data points y i and j The distance between ij in Represents the coordinates of the i-th data point in the k-th dimension in the low-dimensional space; Define the difference function S The gradient descent algorithm is used to update the coordinates of low-dimensional data points. The update formula is as follows: in represents the coordinates of the i-th data point in the k-th dimension, η is the learning rate, is the difference function about the low-dimensional data point position y i Gradient The dimensionality reduction process is performed iteratively until the difference function S converges to the preset threshold ε1, or the maximum number of iterations Tmax is reached.

8. The cultural tourism big data visualization system according to claim 7 is characterized in that: The step of dynamically adjusting the weight of the edge based on the degree of association between data objects using a graph attention network includes: For each data point y in the low-dimensional data i Mapped to a node v in the network graph i , where i = 1, 2, ..., N, and N is the total number of data points; Each node v i Represents a specific cultural and tourism data object; The edge between nodes ij Represents node v i With node v j The correlation between The edge construction formula is as follows: e ij =f(y i ,y j ), where f(y i ,y j ) is the measurement node v i and node v j A function of the similarity between According to the degree of association between nodes, a weight w is assigned to each edge ij Among them, σ is a regularization parameter that controls the smoothness of the weight; Calculate the attention coefficient α between nodes ij Where a is the learned weight vector, T represents the transpose, and N(i) represents the node v i The neighbor set of , || represents the vector connection operation, and LeakyReLU is the activation function; According to the attention coefficient α ij , node v i Update its feature y i , the new node feature is obtained by weighted summation of the features of neighboring nodes: where y i′ is the new node feature, and σ is the activation function.

9. The cultural tourism big data visualization system according to claim 8, characterized in that: Determining the distance between the nodes based on the weight of the edge includes: The edge weight and the distance between nodes d ij The mapping is done by the following formula: Here, ∈ is a constant used to avoid division by zero errors.

10. The cultural tourism big data visualization system according to claim 9, characterized in that: The using the second algorithm to adjust the relative positions between the nodes to obtain the first node layout comprises: The magnitude of attraction is defined as: F attr,ij =k1·w ij ·log(d ij ) Where k1 is a constant that controls the strength of attraction, log(d ij ) is a logarithmic function that is used to balance the rate at which the attractive force changes with distance; The magnitude of the repulsive force is defined as: Among them, k2 is a constant that controls the strength of the repulsive force; Total force between nodes F ij It is the combination of attractive and repulsive forces and is expressed as: F ij =F attr,ij +F repel,ij The update formula of the node position is as follows: in, is the new coordinate, Δt is the time step, m i is the mass of the node.

Images