Supplier portrait visual display method and system based on knowledge graph

Abstract

The invention discloses a supplier portrait visual display method and system based on a knowledge graph, and relates to the technical field of data processing, and the method comprises the following steps: executing information entropy driven compression on a pre-constructed supplier knowledge graph through a graph information bottleneck algorithm, and generating a supplier compressed sub-graph; obtaining a supplier feature set in the supplier compressed sub-graph by positioning a target node; screening a target portrait template from a portrait template library according to the supplier feature set; inputting the supplier compression sub-graph, the supplier feature set and the target portrait template into a dual-channel collaborative embedding architecture to generate a supplier portrait embedding vector; and associating the supplier portrait embedding vector with the target node, and performing visual display. The method is used for solving the problems that in an existing method, supplier feature expression is insufficient, a portrait template lacks adaptability, and visualization result interpretation is insufficient.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a method and system for visualizing supplier portraits based on a knowledge graph. Background Art

[0002] As supply chain networks continue to expand and supplier types become increasingly diverse, companies are facing challenges in procurement and management, such as complex supplier evaluation dimensions, delayed information updates, and insufficient visualization. To achieve refined supplier management and risk control, supplier profiling methods based on knowledge graphs have become a research hotspot. These methods can depict multidimensional relationships between suppliers through graph structures and provide support for supplier evaluation and decision-making.

[0003] However, existing supplier profiling methods still have the following shortcomings: On the one hand, in knowledge graph data processing, most methods only rely on the original full graph for modeling, and fail to effectively distinguish the amount of information of different nodes and edges, resulting in a large graph scale and excessive redundant information, which not only affects the computational efficiency, but also makes the extracted supplier features noisy, making it difficult to accurately characterize the characteristics of the target supplier; On the other hand, in portrait generation and display, most existing methods rely on fixed templates or simple feature splicing, and lack a dynamic matching and collaborative embedding mechanism between supplier features and portrait templates, resulting in the final generated supplier portrait being out of touch with business needs, insufficient interpretability, and difficulty in truly reflecting the supplier's comprehensive capabilities and industry characteristics at the visualization level. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a supplier portrait visualization display method and system based on a knowledge graph, which solves the problems of insufficient expression of supplier characteristics, lack of adaptability of portrait templates and insufficient interpretability of visualization results by introducing a graph information bottleneck algorithm, a portrait template matching mechanism and a dual-channel collaborative embedding architecture.

[0005] To achieve the above object, the present invention provides the following technical solutions: A supplier portrait visualization display method based on a knowledge graph includes the following steps: performing information entropy-driven compression on a pre-built supplier knowledge graph through a graph information bottleneck algorithm to generate a supplier compressed subgraph; obtaining a supplier feature set by locating a target node in the supplier compressed subgraph; screening a target portrait template from a portrait template library based on the supplier feature set; inputting the supplier compressed subgraph, the supplier feature set, and the target portrait template into a dual-channel collaborative embedding architecture to generate a supplier portrait embedding vector; and associating the supplier portrait embedding vector with the target node for visualization.

[0006] In a preferred embodiment, the information entropy-driven compression of the pre-constructed supplier knowledge graph is performed through the graph information bottleneck algorithm to generate a supplier compressed subgraph, specifically: the attribute fields and structural connectivity of all nodes in the supplier knowledge graph are obtained; for each node, the attribute information entropy value is calculated according to the probability distribution of the attribute field value category, and the structural information entropy value is calculated according to the structural connectivity; the attribute information entropy value and the structural information entropy value are weightedly combined to obtain the comprehensive information entropy value of the node; the screening threshold is determined according to the statistical distribution of the comprehensive information entropy values ​​of all nodes; nodes with comprehensive information entropy values ​​greater than or equal to the screening threshold are screened to obtain a set of high-information nodes; an edge selection strategy is executed on the supplier knowledge graph based on the set of high-information nodes to obtain a supplier compressed subgraph.

[0007] In a preferred embodiment, the edge selection strategy is specifically as follows: from the edge set of the supplier knowledge graph, edges whose end nodes both belong to the high-information node set are screened to obtain a retained edge set; multiple edges in the retained edge set that meet a preset similarity threshold or type consistency condition are merged into a single weighted edge, wherein, if the edge to be merged is a non-weighted edge, a default weight value is assigned, and the weights of the edges to be merged are accumulated or weighted averaged to obtain the weight value of a single weighted edge.

[0008] In a preferred embodiment, the supplier feature set is obtained by locating the target node in the supplier compression subgraph, specifically: locating the target node in the supplier compression subgraph, and retrieving the neighborhood nodes of the target node; obtaining the structural feature vectors of the target node and the neighborhood nodes based on the topological structure of the supplier compression subgraph, and aggregating the structural feature vectors of the target node and the neighborhood nodes to obtain a set of structural feature vectors; extracting the behavioral data fields corresponding to the target node and the neighborhood nodes from the supplier data, performing weighted aggregation on the behavioral data fields, and generating a behavioral feature vector; aligning and fusing the structural feature vector set and the behavioral feature vector according to the preset field index order, and combining them with the node classification label of the target node to generate a supplier feature set.

[0009] In a preferred embodiment, the target portrait template is screened from the portrait template library based on the supplier feature set, specifically: several candidate portrait templates are loaded from the portrait template library, and the structural similarity, behavioral similarity and label similarity between the supplier feature set and each candidate portrait template are calculated; the structural similarity, behavioral similarity and label similarity between the supplier feature set and each candidate portrait template are weightedly summed to obtain a multimodal similarity score for each candidate portrait template; the candidate portrait template with the highest multimodal similarity score is selected as the target portrait template; the target portrait template is cropped and field processed to generate an optimized target portrait template.

[0010] In a preferred embodiment, the target portrait template is cropped and field processed, specifically: based on the node classification label of the target node, the fields in the target portrait template that are not related to the industry to which the target node belongs are deleted; the fields that exist in the supplier feature set but are missing in the target portrait template are added to the target portrait template; the fields that exist in the target portrait template but are missing in the supplier feature set are deleted from the target portrait template; and the arrangement order of the fields in the target portrait template is adjusted according to the preset field priority.

[0011] In a preferred embodiment, the dual-channel collaborative embedding architecture includes a structural feature processing channel, a behavioral feature processing channel and a collaborative fusion unit.

[0012] In a preferred embodiment, the supplier compressed subgraph, supplier feature set and target portrait template are input into a dual-channel collaborative embedding architecture to generate a supplier portrait embedding vector, specifically: in the structural feature processing channel, an adjacency relationship is constructed based on the supplier compressed subgraph, and graph convolution operations are performed on the target node and the neighborhood nodes to obtain a structural semantic vector; in the behavioral feature processing channel, multi-layer perceptron modeling is performed on the behavioral feature vector in the supplier feature set to obtain a behavioral semantic vector; in the fusion collaborative unit, based on the field order and field type of the target portrait template, the structural semantic vector and the behavioral semantic vector are respectively mapped to the template field space, and alignment and weighted fusion are performed to generate a supplier portrait embedding vector.

[0013] In a preferred embodiment, the supplier portrait embedding vector is associated with the target node by performing normalization processing on the supplier portrait embedding vector, and mapping it to a two-dimensional or three-dimensional coordinate space through a dimensionality reduction projection algorithm to obtain a mapping coordinate set; binding the mapping coordinate set to the target node, and generating a corresponding visualization node in the knowledge graph visualization interface.

[0014] The supplier portrait visualization display system based on knowledge graph includes: a graph compression module, which is used to perform information entropy-driven compression on the pre-built supplier knowledge graph through the graph information bottleneck algorithm to generate a supplier compressed subgraph; a supplier feature construction module, which is used to obtain the supplier feature set by locating the target node in the supplier compressed subgraph; a portrait template selection module, which is used to filter the target portrait template from the portrait template library according to the supplier feature set; a dual-channel collaborative embedding module, which is used to input the supplier compressed subgraph, the supplier feature set and the target portrait template into the dual-channel collaborative embedding architecture to generate a supplier portrait embedding vector; a visualization module, which is used to associate the supplier portrait embedding vector with the target node for visual display.

[0015] The technical effects and advantages of the supplier portrait visualization display method and system based on knowledge graph of the present invention are as follows: This embodiment introduces a graph information bottleneck algorithm to perform information entropy-driven compression on the pre-built supplier knowledge graph, which can effectively filter out low-value redundant nodes and edge structures, highlight key suppliers and their related factors, and reduce the interference of feature noise on the portrait results, thereby improving the accuracy of feature extraction and the stability of portrait modeling; by combining the dual-channel collaborative embedding architecture of structural features and behavioral features in the supplier compression subgraph, and introducing the field order and type constraints of the target portrait template for collaborative alignment and weighted fusion, it can accurately portray the structural relationship and business behavior characteristics of the supplier, and improve the integrity and interpretability of the portrait results; by introducing a multimodal similarity scoring mechanism in the template selection stage, the structure, behavior and label information are integrated and evaluated, which can ensure a high degree of match between the selected template and the target supplier, and realize dynamic adaptation of the template and actual features in the cropping and field processing links to avoid interference from irrelevant information. As a result, the problems of insufficient expression of supplier features, lack of adaptability of portrait templates, and insufficient interpretability of visualization results in the existing technology are effectively solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of the method for visualizing supplier portraits based on a knowledge graph provided in an embodiment of the present invention.

[0017] Figure 2 A structural diagram of a dual-channel collaborative embedding architecture for a supplier portrait visualization method based on a knowledge graph provided in an embodiment of the present invention.

[0018] Figure 3 A schematic diagram of the system structure of the knowledge graph-based supplier portrait visualization method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will provide a clear and complete description of 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Example 1, Figure 1 A supplier portrait visualization method based on knowledge graph is proposed, which includes the following steps: S1, performs information entropy-driven compression on the pre-built supplier knowledge graph through the graph information bottleneck algorithm to generate a supplier compressed subgraph; S2, obtains the supplier feature set by locating the target node in the supplier compression subgraph; S3, based on the supplier feature set, select the target portrait template from the portrait template library; S4, inputs the supplier compressed subgraph, supplier feature set and target portrait template into the dual-channel collaborative embedding architecture to generate the supplier portrait embedding vector; S5, associate the supplier portrait embedding vector with the target node for visualization.

[0021] This embodiment introduces a graph information bottleneck algorithm to perform information entropy-driven compression on the pre-built supplier knowledge graph, which can effectively filter out low-value redundant nodes and edge structures, highlight key suppliers and their related factors, and reduce the interference of feature noise on the portrait results, thereby improving the accuracy of feature extraction and the stability of portrait modeling; by combining the dual-channel collaborative embedding architecture of structural features and behavioral features in the supplier compression subgraph, and introducing the field order and type constraints of the target portrait template for collaborative alignment and weighted fusion, it can accurately portray the structural relationship and business behavior characteristics of the supplier, and improve the integrity and interpretability of the portrait results; by introducing a multimodal similarity scoring mechanism in the template selection stage, the structure, behavior and label information are integrated and evaluated, which can ensure a high degree of match between the selected template and the target supplier, and realize dynamic adaptation of the template and actual features in the cropping and field processing links to avoid interference from irrelevant information. As a result, the problems of insufficient expression of supplier features, lack of adaptability of portrait templates, and insufficient interpretability of visualization results in the existing technology are effectively solved.

[0022] S1, performs information entropy-driven compression on the pre-built supplier knowledge graph through the graph information bottleneck algorithm to generate a supplier compressed subgraph.

[0023] In this embodiment, the graph information bottleneck algorithm is used to perform information entropy-driven compression on the pre-built supplier knowledge graph to generate a supplier compressed subgraph, specifically: Obtain the attribute fields and structural connectivity of all nodes in the supplier knowledge graph; For each node, the attribute information entropy value is calculated based on the probability distribution of the attribute field value category, and the structural information entropy value is calculated based on the structural connectivity; The attribute information entropy value and the structure information entropy value are weighted and combined to obtain the comprehensive information entropy value of the node; Determine the screening threshold based on the statistical distribution of the comprehensive information entropy values ​​of all nodes; Filter nodes whose comprehensive information entropy value is greater than or equal to the screening threshold to obtain a set of high-information nodes; An edge selection strategy is performed on the supplier knowledge graph based on a set of high-information nodes to obtain a supplier compressed subgraph.

[0024] It should be noted that the graph information bottleneck algorithm is a graph compression method driven by information entropy. While compressing the scale of the graph structure, it maximizes the retention of core information related to the target task and minimizes redundant and noisy information. Furthermore, for the pre-built supplier knowledge graph, the execution process of the graph information bottleneck algorithm includes: calculating the attribute information entropy and structural information entropy of the node, and weighted combination to obtain the comprehensive information entropy; determining the screening threshold to filter out high-information nodes; executing the edge selection strategy on the supplier knowledge graph to obtain the supplier compressed subgraph; Furthermore, the present invention introduces a graph information bottleneck algorithm to perform information entropy-driven compression processing on the pre-constructed supplier knowledge graph, which can effectively retain key information nodes and important semantic relationships in the knowledge graph while reducing redundant data, thereby reducing the data scale while ensuring information integrity and improving overall processing efficiency; the introduction of this algorithm not only significantly optimizes the quality of input features, but also provides a dense and pure input environment for subsequent portrait template selection and dual-channel collaborative embedding, so that the accuracy of template matching and the expressiveness of the embedding vector are enhanced; on this basis, the visualization display module can present clearer industry clustering relationships and supplier portrait differences based on the optimized embedding results, thereby achieving the comprehensive technical effects of noise reduction, quality improvement and enhanced interpretability, forming support and collaborative optimization for the entire supplier portrait generation process.

[0025] It should be noted that the basic structure of the pre-built supplier knowledge graph includes nodes, edges, and attribute fields, wherein the nodes include suppliers and their related business elements, such as supplier enterprise ontology, industry, geographic location, main products, historical transaction records, financial indicators, risk events, partners, customer reviews, etc. The edges represent semantic relationships between nodes, such as "supplier-product supply", "supplier-industry affiliation", "supplier-transaction record", "supplier-partner", "supplier-risk event", etc.; The attribute fields are specific descriptive data attached to a node or edge, such as a supplier's registered capital, credit rating, years of operation, product category, number of collaborations, default rate, etc.

[0026] It should be noted that the structural connectivity specifically reflects the number of connections, connection strength, and connection diversity between a node and other nodes; Furthermore, structural connectivity can be calculated by one or a combination of the following methods: (1) Degree: the number of edges connected to a node; (2) Weighted degree: When the edges have weights, the weights of the connected edges are accumulated to represent the connection strength of the nodes; (3) Connection diversity index: If a node is simultaneously connected to multiple nodes of different categories, its structural connectivity is relatively higher; (4) Centrality indicators: such as betweenness centrality and eigenvector centrality, which are used to measure the propagation role or influence of a node in the entire knowledge graph.

[0027] It should be noted that the probability distribution of the attribute field value categories refers to the statistical distribution of the possible value categories of the specific attribute field of the node in the supplier knowledge graph, and the probability distribution is obtained through frequency statistics or normalization. For example, the attribute fields include the industry, credit rating, region, and order status. The frequency of occurrence of each value category of the industry in the supplier knowledge graph is counted: manufacturing = 30 times, financial industry = 10 times, service industry = 20 times. The frequency of each category is divided by the total frequency to obtain the probability distribution of the value category of the industry: manufacturing industry 0.5, financial industry 0.167, service industry 0.333.

[0028] It should be noted that the calculated attribute information entropy value is specifically: Get the probability distribution value of a node in the actual value category of a certain attribute field (for example, if the industry to which a node belongs is manufacturing, the probability distribution of manufacturing is 0.5); Substitute the probability distribution value of the node in the actual value category of the attribute field into the information entropy formula:

[0029] in, is the information entropy value of the node in the current attribute field, is the probability distribution value of the node in the actual value category of the attribute field (for example, if the industry to which a node belongs is manufacturing, then the information entropy value of the industry described by the node = -log(0.5)); Calculate the information entropy value of each attribute field of the node and perform weighted summation to obtain the attribute information entropy value of the node.

[0030] It should be noted that the structural information entropy value is calculated based on the structural connectivity, specifically: Normalize the distribution of the node's structural connectivity among all nodes to obtain the node's structural probability value:

[0031] in, For nodes The structural probability value of For nodes The structural connectivity of is the node set of the supplier knowledge graph, For nodes The structural connectivity of It is the sum of the structural connectivity of all nodes in the supplier knowledge graph; Calculate the structural information entropy value of the node based on the structural probability value:

[0032] in, For nodes The structural information entropy value of For nodes The structural probability value of .

[0033] It should be noted that the attribute information entropy value and the structure information entropy value are weightedly combined to obtain the comprehensive information entropy value of the node, which is specifically: Based on the attribute information entropy value and the structure information entropy value, the dynamic weight parameters of the node are defined:

[0034] in, For nodes The dynamic weight parameters of , For nodes The structural information entropy value of For nodes The attribute information entropy value of the node is richer. tends to increase, thereby increasing the weight of attribute information in the comprehensive entropy value; on the contrary, when the structural information of the node is more discriminative, tends to decrease, thereby highlighting the contribution of structural information; The attribute information entropy value and the structure information entropy value are weighted and combined to obtain the comprehensive information entropy value of the node:

[0035] in, For nodes The comprehensive information entropy value of For nodes The dynamic weight parameters, For nodes The structural information entropy value of For nodes The attribute information entropy value of .

[0036] It should be noted that the screening threshold is determined based on the statistical distribution of the comprehensive information entropy values ​​of all nodes, specifically: Perform statistical analysis on the comprehensive information entropy values ​​of all nodes to obtain distribution characteristics, such as mean, standard deviation, quantile, etc., and determine the screening threshold based on the distribution characteristics; Furthermore, the screening threshold may be determined specifically as follows: (1) Dynamic threshold based on mean and standard deviation:

[0037] in, is the screening threshold, is the mean, is the standard deviation, is an adjustment parameter (e.g., 0.5–1.0) used to control the stringency of the screening threshold; (2) Quantile-based ratio threshold: directly take the upper quartile as the screening threshold; (3) Adaptive hybrid strategy: combining mean + standard deviation and quantile methods to dynamically select a better threshold.

[0038] In this embodiment, the edge selection strategy is specifically as follows: From the edge set of the supplier knowledge graph, filter the edges whose end nodes both belong to the high-information node set to obtain the retained edge set; Multiple edges in the retained edge set that meet the preset similarity threshold or type consistency condition are merged into a single weighted edge. If the edge to be merged is a non-weighted edge, a default weight value is assigned, and the weights of the edges to be merged are accumulated or weighted averaged to obtain the weight value of the single weighted edge.

[0039] It should be noted that satisfying the preset similarity threshold means that when the similarity between the attribute features of multiple edges is greater than or equal to the preset threshold, these edges are considered to be semantically redundant and have the potential to be merged, wherein the similarity can be calculated by cosine similarity, Jaccard coefficient or Euclidean distance.

[0040] It should be noted that the type consistency condition means that when the semantic types of multiple edges are the same (for example, they are all "cooperation relationship", "transaction relationship" or "supply relationship"), they are considered to be of consistent type, and further checks are made on the time sequence, data source or credibility before deciding whether to merge them.

[0041] It should be noted that the single weighted edge is specifically: If the edge to be merged is a non-weighted edge, it is assigned a default weight value; When all edges to be merged have weight values, the following method is used according to application requirements: Accumulation method: Applicable to scenarios such as transaction frequency and number of interactions. The weight values ​​of all edges to be merged are directly added together to obtain the weight value of a single weighted edge. Weighted average method: Applicable to scoring or probability indicators. The weight values ​​of all edges to be merged are weighted averaged according to the source credibility to obtain the weight value of a single weighted edge. The source credibility is the credibility weight of the edge's data source. For example, the official database source credibility = 0.9, and the third-party industry data source credibility = 0.7.

[0042] S2, obtains the supplier feature set by locating the target node in the supplier compression subgraph.

[0043] In this embodiment, the supplier feature set is obtained by locating the target node in the supplier compression subgraph, specifically: Locate the target node in the supplier compressed subgraph and retrieve the neighboring nodes of the target node; Obtaining the structural feature vectors of the target node and its neighboring nodes based on the topological structure of the supplier compressed subgraph, and aggregating the structural feature vectors of the target node and its neighboring nodes to obtain a set of structural feature vectors; Extract the behavior data fields corresponding to the target node and neighboring nodes from the supplier data, perform weighted aggregation on the behavior data fields, and generate a behavior feature vector; According to the preset field index order, the structural feature vector set and the behavioral feature vector are aligned and fused, and combined with the node classification label of the target node to generate the supplier feature set.

[0044] It should be noted that the target node refers to the node corresponding to the target supplier that needs to be visualized. The user can locate the target node in the supplier compression subgraph by entering the unique identifier of the target supplier.

[0045] It should be noted that the neighboring nodes mainly refer to the first-order neighboring nodes of the target node, and can also be extended to second-order or multi-order neighboring nodes.

[0046] It should be noted that the obtained structural feature vector set is specifically: Based on the topological structure of the supplier compressed subgraph, an adjacency matrix is ​​established; For the target node and its neighboring nodes, basic structural attributes are extracted as the initial structural feature vector. The basic structural attributes include node degree, weighted degree, local clustering coefficient, centrality index of the node in the graph, and node type code. The initial structural feature vectors of the target node and neighboring nodes are input into the adjacency matrix for feature propagation. During the propagation process, the features of the neighboring nodes are attenuated or amplified according to the edge weights and then transmitted to the target node. Perform aggregation operations on the neighborhood feature vectors received by the target node. The aggregation method can be weighted average, maximum pooling, or attention-weighted summation; Finally, the structural feature vectors of the target node and its neighboring nodes are combined to form a set of structural feature vectors.

[0047] It should be noted that the behavioral data field refers to data that reflects the supplier's actual dynamic performance and interactive behavior, including transaction data, contract and cooperation records, credit and compliance behavior, dynamic interactive behavior and industry-related dynamic indicators.

[0048] It should be noted that the structure feature vector set and the behavior feature vector set are aligned and fused according to the preset field index order, and combined with the node classification label of the target node to generate the supplier feature set, specifically: First, the field index order is predefined. The field index order can be set according to the field arrangement standards or industry specifications in the supplier portrait template library. For example, the field index order is: basic attribute field, transaction behavior field, financial risk field, performance capability field, etc.

[0049] Each component in the set of structural feature vectors corresponding to the target node and neighboring nodes is aligned with the corresponding component in the behavioral feature vector in the preset field index order to ensure that the features from different sources are consistent in semantics and position.

[0050] After alignment is completed, the structural features and behavioral features under the same index are fused. The fusion method can be vector concatenation, weighted summation, or linear combination based on correlation coefficients to generate a comprehensive feature vector corresponding to the field index order.

[0051] Finally, the above comprehensive feature vector is combined with the node classification label of the target node to form the supplier feature set.

[0052] S3, based on the supplier feature set, select the target portrait template from the portrait template library.

[0053] In this embodiment, the target portrait template is selected from the portrait template library based on the supplier feature set, specifically: Load several candidate portrait templates from the portrait template library and calculate the structural similarity, behavioral similarity, and label similarity between the supplier feature set and each candidate portrait template; The structural similarity, behavioral similarity, and label similarity between the supplier feature set and each candidate portrait template are weighted and summed to obtain the multimodal similarity score of each candidate portrait template; Select the candidate portrait template with the highest multimodal similarity score as the target portrait template; The target portrait template is cropped and fields are processed to generate an optimized target portrait template.

[0054] It should be noted that the sources of the portrait template library include: first, based on the experience of industry experts and business rules, manually designed standard portrait templates suitable for different industry scenarios, such as supplier portrait templates in the manufacturing, logistics or financial services industries; second, through clustering analysis and statistical modeling of historical supplier data, representative field combinations and feature patterns are automatically mined and solidified into candidate portrait templates; third, enterprise users are supported to expand and customize portrait templates according to business needs during actual use, and newly added templates can also be included in the portrait template library.

[0055] It should be noted that the candidate portrait template is composed of several portrait fields, and each portrait field includes three elements: field name, field type and field weight. Among them, the field name is used to identify the business meaning corresponding to the field, such as "basic information of the enterprise", "financial status", "contract performance", "logistics performance efficiency", "industry classification label", etc.; the field type is used to distinguish the data source or feature category carried by the field, such as structural feature field, behavioral feature field, industry classification label; the field weight is used to indicate the importance of the field in the template. The system can perform differentiated processing based on the field weight when calculating multimodal similarity.

[0056] It should be noted that the structural similarity, behavioral similarity, and label similarity between the supplier feature set and each candidate portrait template are calculated as follows: Extract the template structure vector set, template behavior vector set, and industry classification label of each candidate portrait template through NLP; For each candidate portrait template, a vector similarity metric, such as cosine similarity, is used to calculate the structural similarity between the template structure vector set and the structure feature vector set, and the behavioral similarity between the template behavior vector set and the behavior feature vector set. For label similarity, if the node classification similarity and the industry classification label are the same, the label similarity is set to 1. If they are different, the label similarity can be calculated based on the industry hierarchy tree:

[0057] in, is the label similarity, Classify labels for nodes, For industry classification labels, for and The depth of the common ancestor node in the industry hierarchy tree, e.g. = "Manufacturing > Electronic Components > Semiconductors", = "Manufacturing > Electronic Components > Circuit Boards", the common ancestor of the two is "Manufacturing > Electronic Components", is 2; for Depth in the industry hierarchy tree, for Depth in the industry hierarchy tree.

[0058] It should be noted that, in the weighted summation, the weighted coefficients of structural similarity, behavioral similarity and label similarity can be set through expert experience, learning optimization based on historical data or dynamic adaptive allocation.

[0059] In this embodiment, the target portrait template is cropped and field processed as follows: Based on the node classification label of the target node, delete the fields in the target portrait template that are not related to the industry to which the target node belongs; Add the fields that exist in the supplier feature set but are missing in the target profile template to the target profile template; Delete the fields that exist in the target profile template but are missing from the supplier feature set from the target profile template; Adjust the order of fields in the target profile template according to the preset field priority.

[0060] It should be noted that the preset field priorities can be set based on industry standards or general specifications, or they can be set based on the needs of specific business scenarios. Furthermore, the preset field priorities can be manually configured by experts during system initialization, or dynamically generated through statistical analysis of historical usage data. For example, the higher the frequency of queries or usage in the portrait display for a field, the higher its priority will be.

[0061] S4, the supplier compressed subgraph, supplier feature set and target portrait template are input into the dual-channel collaborative embedding architecture to generate the supplier portrait embedding vector.

[0062] In this embodiment, the dual-channel collaborative embedding architecture includes a structural feature processing channel, a behavioral feature processing channel and a collaborative fusion unit.

[0063] Figure 2 A structural diagram of the dual-channel collaborative embedding architecture of the supplier portrait visualization display method based on knowledge graph is given.

[0064] It should be noted that the two parallel channels in the dual-channel collaborative embedding architecture extract the supplier's relationship structure information and business behavior information respectively, making the feature expression more targeted and complete. The target portrait template is introduced as the fusion reference standard to ensure that the features from different sources can maintain consistency with the portrait display requirements during fusion. The template field space constraint is introduced in the fusion stage to ensure that the embedding result can directly correspond to the portrait display field, shortening the semantic distance from the embedding vector to the visual display.

[0065] In this embodiment, the supplier compressed subgraph, supplier feature set, and target portrait template are input into a dual-channel collaborative embedding architecture to generate a supplier portrait embedding vector, specifically: In the structural feature processing channel, the adjacency relationship is constructed based on the supplier compressed subgraph, and graph convolution operations are performed on the target node and neighboring nodes to obtain the structural semantic vector; In the behavior feature processing channel, multi-layer perceptron modeling is performed on the behavior feature vectors in the supplier feature set to obtain the behavior semantic vectors; In the fusion collaboration unit, based on the field order and field type of the target portrait template, the structural semantic vector and behavioral semantic vector are mapped to the template field space respectively, and aligned and weighted fused to generate the supplier portrait embedding vector.

[0066] S5, associate the supplier portrait embedding vector with the target node for visualization.

[0067] In this embodiment, the supplier portrait embedding vector is associated with the target node, specifically: Normalize the supplier portrait embedding vector and map it to a two-dimensional or three-dimensional coordinate space using a dimensionality reduction projection algorithm to obtain a mapping coordinate set. Bind the mapping coordinate set to the target node and generate the corresponding visualization node in the knowledge graph visualization interface.

[0068] It should be noted that the dimensionality reduction projection algorithm includes but is not limited to the principal component analysis (PCA) algorithm, the t-distributed stochastic neighbor embedding (t-SNE) algorithm, and the unified manifold approximation and projection (UMAP) algorithm.

[0069] Example 2, Figure 3 The present invention provides a system for visualizing supplier portraits based on a knowledge graph, including: The graph compression module is used to perform information entropy-driven compression on the pre-built supplier knowledge graph using the graph information bottleneck algorithm to generate a supplier compressed subgraph; The supplier feature construction module is used to obtain the supplier feature set by locating the target node in the supplier compression subgraph; The portrait template selection module is used to select the target portrait template from the portrait template library based on the supplier feature set; The dual-channel collaborative embedding module is used to input the supplier compressed subgraph, supplier feature set and target portrait template into the dual-channel collaborative embedding architecture to generate the supplier portrait embedding vector; The visualization module is used to associate the supplier portrait embedding vector with the target node for visual display.

[0070] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0071] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0072] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0073] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0074] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0075] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A supplier portrait visualization display method based on knowledge graph, characterized by: The following steps are involved: The pre-built supplier knowledge graph is compressed by information entropy driven using the graph information bottleneck algorithm to generate a supplier compressed subgraph. Obtain the supplier feature set by locating the target node in the supplier compression subgraph; Filter target portrait templates from the portrait template library based on the supplier feature set; The supplier compressed subgraph, supplier feature set, and target profile template are input into a dual-channel collaborative embedding architecture to generate a supplier profile embedding vector. Associate the supplier portrait embedding vector with the target node for visualization.

2. The supplier portrait visualization display method based on knowledge graph according to claim 1 is characterized in that: The graph information bottleneck algorithm is used to perform information entropy-driven compression on the pre-built supplier knowledge graph to generate a supplier compressed subgraph, specifically: Obtain the attribute fields and structural connectivity of all nodes in the supplier knowledge graph; For each node, the attribute information entropy value is calculated based on the probability distribution of the attribute field value category, and the structural information entropy value is calculated based on the structural connectivity; The attribute information entropy value and the structure information entropy value are weighted and combined to obtain the comprehensive information entropy value of the node; Determine the screening threshold based on the statistical distribution of the comprehensive information entropy values ​​of all nodes; Filter nodes whose comprehensive information entropy value is greater than or equal to the screening threshold to obtain a set of high-information nodes; An edge selection strategy is performed on the supplier knowledge graph based on a set of high-information nodes to obtain a supplier compressed subgraph.

3. The supplier portrait visualization display method based on knowledge graph according to claim 2 is characterized in that: The edge selection strategy is specifically as follows: From the edge set of the supplier knowledge graph, filter the edges whose end nodes both belong to the high-information node set to obtain the retained edge set; Multiple edges in the retained edge set that meet the preset similarity threshold or type consistency condition are merged into a single weighted edge. If the edge to be merged is a non-weighted edge, a default weight value is assigned, and the weights of the edges to be merged are accumulated or weighted averaged to obtain the weight value of the single weighted edge.

4. The supplier portrait visualization display method based on knowledge graph according to claim 3 is characterized in that: The supplier feature set is obtained by locating the target node in the supplier compression subgraph, specifically: Locate the target node in the supplier compressed subgraph and retrieve the neighboring nodes of the target node; Obtaining the structural feature vectors of the target node and its neighboring nodes based on the topological structure of the supplier compressed subgraph, and aggregating the structural feature vectors of the target node and its neighboring nodes to obtain a set of structural feature vectors; Extract the behavior data fields corresponding to the target node and neighboring nodes from the supplier data, perform weighted aggregation on the behavior data fields, and generate a behavior feature vector; According to the preset field index order, the structural feature vector set and the behavioral feature vector are aligned and fused, and combined with the node classification label of the target node to generate the supplier feature set.

5. The supplier portrait visualization display method based on knowledge graph according to claim 4 is characterized in that: The target portrait template is selected from the portrait template library based on the supplier feature set, specifically: Load several candidate portrait templates from the portrait template library and calculate the structural similarity, behavioral similarity, and label similarity between the supplier feature set and each candidate portrait template; The structural similarity, behavioral similarity, and label similarity between the supplier feature set and each candidate portrait template are weighted and summed to obtain the multimodal similarity score of each candidate portrait template; Select the candidate portrait template with the highest multimodal similarity score as the target portrait template; The target portrait template is cropped and fields are processed to generate an optimized target portrait template.

6. The supplier portrait visualization display method based on knowledge graph according to claim 5 is characterized in that: The target portrait template is cropped and field processed as follows: Based on the node classification label of the target node, delete the fields in the target portrait template that are not related to the industry to which the target node belongs; Add the fields that exist in the supplier feature set but are missing in the target profile template to the target profile template; Delete the fields that exist in the target profile template but are missing from the supplier feature set from the target profile template; Adjust the order of fields in the target profile template according to the preset field priority.

7. The supplier portrait visualization display method based on knowledge graph according to claim 6 is characterized in that: The dual-channel collaborative embedding architecture includes a structural feature processing channel, a behavioral feature processing channel and a collaborative fusion unit.

8. The supplier portrait visualization display method based on knowledge graph according to claim 7 is characterized in that: The supplier compressed subgraph, supplier feature set and target portrait template are input into the dual-channel collaborative embedding architecture to generate the supplier portrait embedding vector, specifically: In the structural feature processing channel, the adjacency relationship is constructed based on the supplier compressed subgraph, and graph convolution operations are performed on the target node and neighboring nodes to obtain the structural semantic vector; In the behavior feature processing channel, multi-layer perceptron modeling is performed on the behavior feature vectors in the supplier feature set to obtain the behavior semantic vectors; In the fusion collaboration unit, based on the field order and field type of the target portrait template, the structural semantic vector and behavioral semantic vector are mapped to the template field space respectively, and aligned and weighted fused to generate the supplier portrait embedding vector.

9. The supplier portrait visualization display method based on knowledge graph according to claim 8 is characterized in that: The supplier portrait embedding vector is associated with the target node, specifically: Normalize the supplier portrait embedding vector and map it to a two-dimensional or three-dimensional coordinate space using a dimensionality reduction projection algorithm to obtain a mapping coordinate set. Bind the mapping coordinate set to the target node and generate the corresponding visualization node in the knowledge graph visualization interface.

10. A system using the method for visualizing supplier portraits based on a knowledge graph according to any one of claims 1 to 9, comprising: The graph compression module is used to perform information entropy-driven compression on the pre-built supplier knowledge graph using the graph information bottleneck algorithm to generate a supplier compressed subgraph; The supplier feature construction module is used to obtain the supplier feature set by locating the target node in the supplier compression subgraph; The portrait template selection module is used to filter the target portrait template from the portrait template library based on the supplier feature set; The dual-channel collaborative embedding module is used to input the supplier compressed subgraph, supplier feature set and target portrait template into the dual-channel collaborative embedding architecture to generate the supplier portrait embedding vector; The visualization module is used to associate the supplier portrait embedding vector with the target node for visual display.

Images