Data annotation method, device and storage medium based on graph structure and community discovery

Abstract

The present application discloses a data annotation method, device and storage medium based on graph structure and community discovery, which belongs to the field of data processing. The present application obtains customer transactions, product attributes, behavior trajectories and environmental data, and constructs a heterogeneous graph network including a basic physical layer (transaction association and product attribute mapping), a behavioral semantic layer (behavioral pattern and semantic association), and an environmental association layer (dynamic impact of the environment). A community discovery algorithm is used to mine customer groups with cross-departmental business value, and a three-level labeling system is constructed to quantify basic value attributes and fluctuation coefficients, behavioral patterns and product preferences, and dynamic trajectory characteristics. This achieves a unified understanding of customer behavior, in-depth mining of multi-dimensional customer relationships, dynamic evolution capture of customer value, and hierarchical customer cognition construction, solving the problems of data silos caused by task orientation in traditional methods, the constraints of dynamic value mining caused by a single data dimension, and the restrictions of hierarchical cognition construction caused by the flattening of the labeling system.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a data annotation method, device, and storage medium based on graph structure and community discovery. Background Art

[0002] In the digital transformation of the retail industry, customer data labeling, as a key step in data preprocessing, is crucial for building refined customer profiles and supporting intelligent marketing decisions. Traditional customer data labeling methods in retail scenarios primarily focus on specific business needs, such as product recommendation systems and membership management systems, assigning static labels (such as "high-value customers" or "preferred by maternal and infant products") to customers manually or through rule engines. For example, patent application publication number CN118535775A, entitled "A Retail Shopping Guide Recommendation Method and System Based on User Profile Labels," discloses a method for combining real-time consumer demand and historical consumer preferences to construct profile labels and provide personalized product recommendations to users. Another example is patent application publication number CN115619454A, entitled "A Membership Management System and Method Based on a Membership Middle Platform," which discloses a method for adding user tags to members through multi-channel data classification and generating personalized marketing campaigns based on these tags. While these methods meet the customer classification needs of basic business scenarios to a certain extent, their limitations are becoming increasingly prominent as retail scenarios evolve towards omnichannel and personalized approaches:

[0003] (1) Excessive task orientation leads to data silos: Retail business naturally involves the collaboration of multiple departments (such as procurement, inventory, marketing, and after-sales), and the problem of data fragmentation between departments is particularly prominent. The traditional labeling process is highly dependent on specific business goals (such as promotional activity response prediction), and different business departments often independently build labeling systems to meet their specific business needs. For example, the marketing department may only focus on transaction data and label "high-value customers" based on transaction data; while the risk control department may only focus on behavioral data and label "Double 11 high-spending users" based on behavioral data. This approach of independently building a labeling system based on departmental KPIs exacerbates the phenomenon of data silos, causing the same customer to be labeled with fragmented labels in different retail departments, making it difficult to form a unified customer cognition across departments, thereby affecting the overall operational efficiency and customer experience of retail companies;

[0004] (2) The single dimension of data restricts value mining: Traditional methods over-rely on structured data such as transaction amount and purchase frequency, ignoring the associated value of unstructured information such as customer behavior trajectory in physical stores (such as heat map of traffic flow), online browsing path (such as product detail page jump logic) and environmental factors (such as holiday promotion cycle, weather changes). As a result, the annotation results can only reflect the static cross-section of customer value and cannot capture dynamic evolution characteristics such as "adjustment of shopping categories due to seasonal changes";

[0005] (3) Flat label system limits cognitive depth: Existing technologies usually adopt a flat structure of "single-layer label + weight value" (such as "maternal and infant product preference: 0.8"), which can neither establish a hierarchical cognitive framework such as "high-value customers → maternal and infant product preference → milk powder related purchases" nor quantify temporal characteristics such as "purchase cycle volatility" and "activity response attenuation rate", resulting in a lack of data support for precision marketing in complex scenarios (such as life cycle value prediction).

[0006] Currently, no effective solution has been proposed to the technical problems existing in the above-mentioned existing technologies, such as the traditional customer data labeling methods being too task-oriented, leading to data silos, the single data dimension restricting dynamic value mining, and the flat label system limiting hierarchical cognitive construction. Summary of the Invention

[0007] The embodiments of the present disclosure provide a data labeling method, device and storage medium based on graph structure and community discovery, so as to at least solve the technical problems existing in the prior art of traditional customer data labeling methods, such as data silos caused by excessive task orientation, single data dimension restricting dynamic value mining, and flat label system limiting hierarchical cognitive construction.

[0008] According to one aspect of an embodiment of the present disclosure, a data annotation method based on graph structure and community discovery is provided, which is applied to a data annotation platform in a retail scenario, wherein the data annotation method comprises: obtaining multi-dimensional data of a customer group to be labeled; wherein the multi-dimensional data comprises customer transaction data, product attribute data, customer behavior trajectory data, and environmental data; constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network comprises a basic physical layer, a behavioral semantic layer, and an environmental association layer, wherein the basic physical layer is used to characterize the transaction association relationship between customers and products and the physical attributes of the products, the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products, and the environmental association layer is used to characterize the external environmental factors. The method comprises the following steps: the first step is to determine the dynamic influence of factors on the interaction between commodities and customers; and the second step is to divide the customer group into multiple customer communities through a community discovery algorithm based on the heterogeneous graph network, and generate a hierarchical label tree for each customer community; wherein the hierarchical label tree includes a first-level label, a second-level label and a third-level label, and the first-level label is constructed based on the interaction mode of the basic physical layer and the environmental association layer, and serves as the top-level node of the hierarchical label tree; the second-level label is constructed based on the potential behavior mode of customers in the customer community in the behavior semantic layer, and serves as the middle-level node of the hierarchical label tree; the third-level label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom-level node of the hierarchical label tree.

[0009] According to another aspect of an embodiment of the present disclosure, a storage medium is further provided, the storage medium including a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0010] According to another aspect of the embodiment of the present disclosure, a data labeling device based on graph structure and community discovery is also provided, including: a data acquisition module for acquiring multi-dimensional data of customer groups to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; a graph construction module for constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer and an environmental association layer, wherein the basic physical layer is used to characterize the transaction association relationship between customers and products and the physical attributes of the products, the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products, and the environmental association layer is used to characterize the influence of external environmental factors on the relationship between products and customers. The dynamic impact of customer interaction behavior; and a labeling module, which is used to divide the customer group into multiple customer communities based on the heterogeneous graph network through a community discovery algorithm, and generate a hierarchical label tree for each customer community; wherein the hierarchical label tree includes a first-level label, a second-level label and a third-level label, the first-level label is constructed based on the interaction mode of the basic physical layer and the environmental association layer, and serves as the top-level node of the hierarchical label tree; the second-level label is constructed based on the potential behavior mode of customers in the customer community in the behavior semantic layer, and serves as the middle-level node of the hierarchical label tree; the third-level label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom-level node of the hierarchical label tree.

[0011] According to another aspect of the embodiment of the present disclosure, a data labeling device based on graph structure and community discovery is also provided, including: a processor; and a memory, connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining multi-dimensional data of the customer group to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer and an environmental association layer, wherein the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products, the behavioral semantic layer is used to characterize the customer's potential behavior pattern and the semantic association characteristics between products, and the environmental association layer is used to Characterize the dynamic impact of external environmental factors on the interaction behavior between goods and customers; and based on the heterogeneous graph network, divide the customer group into multiple customer communities through a community discovery algorithm, and generate a hierarchical label tree for each customer community; wherein, the hierarchical label tree includes a first-level label, a second-level label and a third-level label, the first-level label is constructed based on the interaction pattern of the basic physical layer and the environmental association layer, and serves as the top-level node of the hierarchical label tree; the second-level label is constructed based on the potential behavior pattern of customers in the customer community in the behavior semantic layer, and serves as the middle-level node of the hierarchical label tree; the third-level label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom-level node of the hierarchical label tree.

[0012] This application first obtains multi-dimensional data such as customer transaction data, product attribute data, customer behavior trajectory data, and environmental data in retail scenarios, breaking the data fragmentation caused by departmental task differences in the traditional labeling process. Then, by constructing a heterogeneous graph network containing a basic physical layer, a behavioral semantic layer, and an environmental association layer, these data from different sources and modalities are uniformly mapped into a graph structure, expressing the complex relationship between customers, products, and the environment in the form of a graph structure, thereby achieving deep integration of multi-dimensional data and multi-dimensional relationship modeling in the retail scenario, providing a unified data foundation for a comprehensive understanding of customer behavior. Secondly, through a community discovery algorithm, customer groups with cross-business department business value are identified in the heterogeneous graph network. These groups are not defined based on the needs of a single department, but through the cross-layer connection characteristics of the graph network, they naturally map common behavior patterns and value characteristics in different business scenarios, thereby forming a unified cognitive framework for customers. This framework supports cross-business department applications and solves the problem of the same customer being labeled with fragmented labels in the traditional labeling system. Finally, based on the above cross-business department customer data, a hierarchical cognitive framework is established by constructing a hierarchical label tree. This hierarchical tag tree not only reflects the logical relationships between different tags but also, through the construction of multi-level tags, meticulously characterizes the multiple dimensions and levels of customer value. This achieves the technical benefits of a unified understanding of customer behavior, in-depth exploration of multidimensional customer relationships, capturing the dynamic evolution of customer value, and building hierarchical customer cognition. This addresses the technical issues of existing customer data tagging methods, such as their overly task-oriented nature leading to data silos, a single data dimension that restricts dynamic value mining, and a flat tag system that limits hierarchical cognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of this application. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0014] Figure 1 is a hardware structure block diagram of a computing device for implementing the method according to embodiment 1 of the present disclosure;

[0015] Figure 2 1 is a schematic diagram of the hardware architecture of the data annotation system based on graph structure and community discovery according to Example 1 of the present disclosure;

[0016] Figure 3 This is a flowchart of the data annotation method based on graph structure and community discovery according to Example 1 of the present application;

[0017] Figure 4 is a schematic diagram of a data annotation device based on graph structure and community discovery according to Example 2 of the present application;

[0018] Figure 5 This is a schematic diagram of a data annotation device based on graph structure and community discovery according to Example 3 of the present application. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present disclosure.

[0020] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] Example 1

[0022] According to this embodiment, a method embodiment of a data labeling method based on graph structure and community discovery is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0023] The method embodiment provided in this embodiment can be executed in a server or similar computing device. Figure 1 The hardware structure block diagram of a computing device for implementing a data annotation method based on graph structure and community discovery is shown. Figure 1As shown, a computing device may include one or more processors (the processor may include, but is not limited to, a microprocessor (MCU) or a programmable logic device (FPGA) or other processing device), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, transmission device, and input / output interface are connected to the processor via a bus. In addition, it may also include: a display, a keyboard, and a cursor control device connected to the input / output interface. Those skilled in the art will understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0024] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry." The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be fully or partially integrated into any of the other components of the computing device. As discussed in the embodiments of the present disclosure, the data processing circuitry functions as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0025] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the data annotation method based on graph structure and community discovery in the embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the data annotation method based on graph structure and community discovery of the above-mentioned application. The memory may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely located relative to the processor, and these remote memories may be connected to the computing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0026] The transmission device is used to receive or send data via a network. Specific examples of the aforementioned network may include a wireless network provided by a communications provider of the computing device. In one embodiment, the transmission device includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0027] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computing device.

[0028] It should be noted that, in some optional embodiments, the above Figure 1 The computing device shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computing device described above.

[0029] Figure 2 Schematic diagram of the data annotation system based on graph structure and community discovery according to this embodiment. Figure 2 As shown, the system includes a server 300 of the data annotation platform. Each terminal device (e.g., terminal device 200) running a client of the data annotation platform can communicate with the server 300. The server 300 can communicate with a database 400 in the data annotation platform and support reading or storing corresponding data from the database 400.

[0030] A user or business entity 100 with data annotation needs can send a data annotation request containing multi-dimensional data (including customer transaction data, product attribute data, customer behavior trajectory data, and environmental data) to a server 300 via a terminal device (e.g., terminal device 200) running a client of the data annotation platform. Server 300 can cleanse the multi-dimensional data in the request and store it in database 400 for subsequent processing. Furthermore, after server 300 completes the annotation of a customer group, it can store the intermediate data generated during the annotation process and the final annotation results in database 400 for query by the terminal device. This allows the terminal device user to subsequently perform downstream task analysis or model training based on the final annotation results obtained from server 300.

[0031] It should be noted that the server 300 of the data annotation platform can be applied to the hardware structure described above.

[0032] Under the above operating environment, according to the first aspect of this embodiment, a data annotation method based on graph structure and community discovery is provided. Figure 2 The server 300 shown in FIG. Figure 3 A schematic diagram of the process is shown in FIG. Figure 3 As shown, the method includes:

[0033] S302: Acquire multi-dimensional data of the customer group to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data, and environmental data;

[0034] S304: Constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer, and an environmental association layer, wherein the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products; the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products; and the environmental association layer is used to characterize the dynamic impact of external environmental factors on the interaction between products and customers; and

[0035] S306: Based on the heterogeneous graph network, the customer group is divided into multiple customer communities through a community discovery algorithm, and a hierarchical label tree is generated for each customer community; wherein the hierarchical label tree includes a first-level label, a second-level label and a third-level label, the first-level label is constructed based on the interaction mode of the basic physical layer and the environmental association layer, and serves as the top-level node of the hierarchical label tree; the second-level label is constructed based on the potential behavior pattern of customers in the customer community in the behavioral semantic layer, and serves as the middle-level node of the hierarchical label tree; the third-level label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom-level node of the hierarchical label tree.

[0036] In an embodiment of the present invention, when a user or business party in any business department (such as procurement, inventory, marketing, after-sales, etc.) has a need to label a certain customer group, a data labeling request can be sent to the server 300 through a terminal device. The data labeling request includes multi-dimensional data of the customer group, providing data support for subsequent labeling processing. Multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data, and environmental data. Specifically, customer transaction data includes order ID, transaction amount, transaction timestamp, etc. Product attribute data refers to information used to describe product characteristics, performance, function, appearance, specifications, etc. Customer behavior trajectory data includes the length of time spent on the product details page, add-to-cart events, etc. Environmental data includes temperature, customer flow density, holidays, and other information.

[0037] After server 300 obtains multi-dimensional data on customer groups, it needs to construct a heterogeneous graph network based on this data, including a basic physical layer, a behavioral semantic layer, and an environmental association layer. This heterogeneous graph network aims to comprehensively depict the complex relationships between customers, products, and the environment. The specific construction steps are as follows:

[0038] (1) Server 300 first defines the node types in the heterogeneous graph network, such as customer nodes, product nodes, and environment nodes, and determines the attribute information of each node type. For example, a customer node may contain attributes such as customer ID, age, and gender; a product node may contain attributes such as product ID, category, and price.

[0039] (2) Server 300 constructs a basic physical layer based on the transaction relationship between customers and products and the mapping relationship between the physical properties of the products. Specifically, server 300 constructs a basic physical layer for representing the transaction relationship between customers and products and the physical properties of the products based on the node attributes and corresponding edge relationships of customer nodes, product nodes, and region nodes.

[0040] (3) The server 300 constructs a behavioral semantic layer based on the dimension of customer potential behavior patterns and semantic association features between products. Specifically, the server 300 mines customer potential behavior patterns, such as browsing habits and purchasing preferences, by analyzing customer behavior trajectory data, and uses these patterns as node attributes of the behavioral semantic layer. In addition, the server 300 also calculates the semantic association features between products, such as extracting keywords from product descriptions through text analysis, and constructing semantic association edges between products. Afterwards, the server 300 constructs a behavioral semantic layer based on the node attributes and corresponding edge relationships of customer behavior nodes and product combination nodes, customer potential behavior patterns, semantic association features, and semantic association edges between products.

[0041] (4) Server 300 constructs an environmental association layer based on the dynamic impact of external environmental factors on the interaction between products and customers. Specifically, server 300 integrates environmental data, such as weather, holidays, and promotional activities, and analyzes how these factors affect the interaction between customers and products. For example, server 300 can analyze the impact of rainy days on customer visit rates and product sales, and use this impact as the edge weight of the environmental association layer. Subsequently, server 300 constructs the environmental association layer based on the node attributes of the environmental nodes and the corresponding edge relationships and edge weights.

[0042] (5) Server 300 connects the basic physical layer, behavioral semantic layer, and environmental association layer into an overall heterogeneous graph network.

[0043] Through the above approach, server 300 successfully constructed a heterogeneous graph network that comprehensively reflects the complex relationships between customers, products, and the environment, achieving the goal of uniformly mapping multidimensional data into a graph structure. This network not only expresses the complex relationships between customers, products, and the environment, but also supports cross-business department data fusion and multidimensional relationship modeling, providing strong support for subsequent data analysis and mining.

[0044] Next, the server 300 divides the customer group into multiple customer communities based on the heterogeneous graph network using a community discovery algorithm and generates a hierarchical label tree for each customer community. Specifically, the server 300 performs the following operations:

[0045] (1) Server 300 selects a community discovery algorithm suitable for heterogeneous graph networks, such as existing deep learning-based community discovery algorithms (such as GraphSAGE and GAT) or traditional community discovery algorithms (such as Louvain and Leiden algorithms). These algorithms can take into account the characteristics of different node types and edge types in heterogeneous graph networks, thereby more accurately discovering customer communities. Server 300 adjusts algorithm parameters, such as the number of communities and the number of iterations, based on the characteristics of heterogeneous graph networks to optimize the community discovery effect. Through the community discovery algorithm, server 300 divides the customer group into multiple customer communities with internal similarities and external differences.

[0046] (2) For each discovered customer community, the server 300 generates a hierarchical tag tree to comprehensively and deeply describe the characteristics of the customers in the community. Specifically, based on the interaction patterns of the basic physical layer and the environmental association layer, the server 300 extracts the basic value attributes and value fluctuation coefficients of the customer community. These attributes are quantified into first-level tags, which serve as the top-level nodes of the hierarchical tag tree and reflect the overall value characteristics of the customer community. The server 300 analyzes the interaction behavior patterns of customers in the customer community with the products in the behavioral semantic layer, such as purchase preferences, browsing habits, etc. Through cluster analysis or pattern mining, the server 300 generates second-level tags as the middle-level nodes of the hierarchical tag tree to deeply characterize the behavioral characteristics of the customer community. The server 300 captures the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, such as the changing trend of purchase frequency and the evolution of behavioral patterns. These time series characteristics are converted into third-level tags as the bottom-level nodes of the hierarchical tag tree to dynamically reflect the changing characteristics of the customer community.

[0047] Through the above steps, the server 300 not only realizes the community division of customer groups, but also generates a comprehensive and in-depth hierarchical label tree for each customer community as the final labeling result, based on which downstream task analysis or model training can be carried out.

[0048] As described in the background technology, existing customer data labeling methods have defects such as being too task-oriented, leading to data silos, a single data dimension restricting dynamic value mining, and a flat label system limiting hierarchical cognitive construction.

[0049] In view of this, this application first obtains multi-dimensional data such as customer transaction data, product attribute data, customer behavior trajectory data, and environmental data in retail scenarios, breaking the data fragmentation caused by departmental task differences in the traditional labeling process. Then, by constructing a heterogeneous graph network containing a basic physical layer, a behavioral semantic layer, and an environmental association layer, these data from different sources and modalities are uniformly mapped into a graph structure, expressing the complex relationship between customers, products, and the environment in the form of a graph structure, thereby achieving deep integration of multi-dimensional data and multi-dimensional relationship modeling in the retail scenario, providing a unified data foundation for a comprehensive understanding of customer behavior. Secondly, through a community discovery algorithm, customer groups with cross-department business value are identified in the heterogeneous graph network. These groups are not defined based on the needs of a single department, but through the cross-layer connection characteristics of the graph network, they naturally map common behavior patterns and value characteristics in different business scenarios, thereby forming a unified cognitive framework for customers. This framework supports cross-department application and solves the problem of fragmented labels being attached to the same customer in the traditional labeling system. Finally, based on the above cross-department customer data, a hierarchical cognitive framework is established by constructing a hierarchical label tree. This hierarchical tag tree not only reflects the logical relationships between different tags but also, through the construction of multi-level tags, meticulously characterizes the multiple dimensions and levels of customer value. This achieves the technical benefits of a unified understanding of customer behavior, in-depth exploration of multidimensional customer relationships, capturing the dynamic evolution of customer value, and building hierarchical customer cognition. This addresses the technical issues of existing customer data tagging methods, such as their overly task-oriented nature leading to data silos, a single data dimension that restricts dynamic value mining, and a flat tag system that limits hierarchical cognition.

[0050] Optionally, the operation of constructing a heterogeneous graph network based on the multi-dimensional data includes: determining the node attributes of six pre-defined node types based on the multi-dimensional data; wherein the six node types include customer nodes, product nodes, regional nodes, environmental nodes, customer behavior nodes and product combination nodes; analyzing the multi-dimensional data to construct edge relationships between the six node types; constructing a basic physical layer for characterizing the transaction relationship between customers and products and the physical attributes of products based on the node attributes and corresponding edge relationships of the customer nodes, the product nodes and the regional nodes; on top of the basic physical layer, constructing a behavioral semantic layer for characterizing potential customer behavior patterns and semantic association features between products based on the node attributes and corresponding edge relationships of the customer behavior nodes and the product combination nodes; on top of the behavioral semantic layer, constructing an environmental association layer for characterizing the dynamic impact of external environmental factors on the interactive behavior between products and customers based on the node attributes and corresponding edge relationships of the environmental nodes, to obtain a preliminary heterogeneous graph network; and establishing a cross-layer interaction mechanism for the preliminary heterogeneous graph network, and performing multi-granularity attribute fusion processing to generate a final heterogeneous graph network.

[0051] In an embodiment of the present invention, the server 300 can predefine six types of nodes, namely customer nodes, product nodes, regional nodes, environmental nodes, customer behavior nodes, and product combination nodes. The server 300 then analyzes the multi-dimensional data to determine the node attributes of these six types of nodes. Specifically, the server 300 analyzes customer transaction data and extracts basic customer information (such as ID, age, gender, membership, etc.) as the node attributes of the customer node. The server 300 analyzes product attribute data and extracts information such as the product ID, name, category, price, size, and weight as the node attributes of the product node. The server 300 analyzes customer behavior trajectory data and extracts information such as the ID, location, area, and regional sales of the customer's region as the node attributes of the region node. Server 300 analyzes environmental data and extracts environmental characteristics (such as temperature, humidity, customer flow density, promotional event ID, time, discount rate, etc.) as node attributes of environmental nodes. Server 300 analyzes customer behavior trajectory data and extracts customer behavior characteristics (such as click, browse, add-to-cart events, behavior trajectory paths, dwell time, etc.) as node attributes of customer behavior nodes. Server 300 analyzes product attribute data and extracts characteristics of frequently co-occurring product combinations (such as product co-occurrence frequency, complementarity coefficient, and combination recommendation weight) as node attributes of product combination nodes.

[0052] Next, the server 300 needs to further analyze the multi-dimensional data and construct the edge relationship between the above six types of nodes. Specifically, the server 300 further analyzes the customer transaction data and establishes a transaction association edge between the customer and the product. The edge weight is related to, for example, the transaction amount and frequency. The server 300 further analyzes the product attribute data and establishes an attribute association edge between the product and the product. The edge weight is related to, for example, the similarity of the product. The server 300 further analyzes the customer behavior trajectory data and establishes a behavior sequence edge between the customer behavior node and the customer node. The edge weight is related to, for example, the frequency of the behavior or the conversion rate. The server 300 further analyzes the customer transaction data and the product attribute data, extracts the product combination sales data, and then analyzes the product combination sales data to establish a combination association edge between the product combination node and the product node. The edge weight is related to, for example, the contribution of the product to the combination sales. Server 300 jointly analyzes environmental data, customer transaction data, and product attribute data, extracting environmental impact information between the three. Based on this environmental impact information, it constructs environmental impact edges between environmental nodes and customer nodes, as well as environmental impact edges between product nodes. Edge weights are, for example, related to the degree to which environmental factors influence customer behavior or product sales. Server 300 jointly analyzes environmental data, customer transaction data, and product attribute data, extracting seasonal variation information between the three. Based on this seasonal variation information, it constructs environmental adjustment edges between environmental nodes and product combination nodes. Edge weights are, for example, related to the impact of seasonal factors on product combination sales.

[0053] Subsequently, based on the node attributes and corresponding edge relationships of customer, product, and region nodes, server 300 constructs a basic physical layer to characterize the transaction relationships between customers and products, as well as the physical properties of the products. Specifically, server 300 constructs a transaction network between customers and products based on the transaction edge relationships between customer and product nodes, serving as the initial basic physical layer. The weights of the transaction edges can reflect key indicators such as transaction amount and frequency, quantifying the intensity of transactions between customers and products. Server 300 maps the physical properties of product nodes (such as size, color, and material) into the basic physical layer. These properties can be connected to product nodes via attribute edges, forming a mapping network of product physical properties. For example, the size attribute of a product can be represented by a virtual node and connected to a specific product node via attribute edges. Based on the sales edge relationships between region and product nodes, server 300 incorporates the sales characteristics of the region nodes (such as regional sales volume and inventory level) into the basic physical layer. The weights of the sales edges can reflect the sales performance of products in different regions. This constructs a basic physical layer that accurately characterizes the transaction relationships between customers and products, as well as the physical properties of the products.

[0054] Next, server 300 constructs a behavioral semantic layer on top of the basic physical layer, based on the node attributes and corresponding edge relationships of the customer behavior nodes and product combination nodes, to characterize potential customer behavior patterns and semantic associations between products. Specifically, server 300 constructs a network of potential customer behavior patterns on top of the basic physical layer, as a preliminary behavioral semantic layer, based on the behavior sequence edge relationships between customer behavior nodes and product combination nodes. The weights of the behavior sequence edges can reflect the frequency of behavior, conversion rate, or temporal importance, quantifying the strength of the association between customer behavior and product combinations. Server 300 maps the semantic association features of the product combination nodes (such as product co-occurrence frequency and combination recommendation weight) into the behavioral semantic layer. These features can be connected to the product combination nodes via semantic association edges, forming a mapping network of semantic associations between products. For example, a product combination that is frequently purchased together can be represented by a virtual node, connected to the specific product combination node via semantic association edges, thereby reflecting the semantic association features between products. This constructs a behavioral semantic layer that accurately characterizes potential customer behavior patterns and semantic association features between products.

[0055] Next, server 300 constructs an environmental association layer on top of the behavioral semantic layer, based on the node attributes and corresponding edge relationships of the environmental nodes, to characterize the dynamic impact of external environmental factors on product-customer interactions, thereby generating a preliminary heterogeneous graph network. Specifically, based on known environmental node attributes (such as temperature, humidity, and promotional activity identifiers) and pre-set edge relationship rules (such as the strength of the association between "high temperature environment" and "cold beverages"), server 300 maps the impact of environmental factors on customer behavior patterns (such as the rate of change in purchase frequency) into environmental influence edges through weight configuration. It also maps the regulatory effect of environmental factors on product combination sales (such as the boost coefficient of holiday promotions on product combination sales) into environmental regulation edges. Furthermore, on top of the behavioral semantic layer, an environmental association layer is constructed based on the node attributes of the environmental nodes and the corresponding environmental influence edges and environmental regulation edges.

[0056] Finally, server 300 establishes a cross-layer interaction mechanism for the preliminary heterogeneous graph network and performs multi-granular attribute fusion processing to generate the final heterogeneous graph network. Specifically, server 300 implements information exchange between the basic physical layer, behavioral semantic layer, and environmental association layer through pre-set cross-layer connection rules (such as the indirect association path between customer nodes and product combination nodes). It also employs a dynamic weight adjustment strategy (such as real-time adjustment of behavioral sequence edge weights based on environmental changes) to enhance the network's adaptability to complex scenarios. At the attribute fusion level, server 300 performs weighted fusion of node attributes (such as transaction amount and behavior frequency), ultimately generating a heterogeneous graph network capable of cross-layer interaction and multi-granular attribute fusion.

[0057] Through this approach, a heterogeneous graph network was constructed that comprehensively reflects the complex relationships between customers, products, and the environment, achieving the goal of uniformly mapping multidimensional data into a graph structure. This network not only expresses the complex relationships between customers, products, and the environment, but also supports cross-business department data fusion and multidimensional relationship modeling, providing strong support for subsequent data analysis and mining.

[0058] Optionally, the operation of determining the node attributes of the predefined six types of nodes based on the multi-dimensional data includes: extracting customer static features from the customer transaction data as node attributes of the customer node; wherein the customer static features include customer ID, membership level and life cycle stage; extracting product static features from the product attribute data as node attributes of the product node; wherein the product static features include product ID, physical attributes and price range; extracting regional features from the customer behavior trajectory data as node attributes of the regional node; wherein the regional features include sales area coordinates and delivery range; extracting environmental features from the environmental data as node attributes of the environmental node; wherein the environmental features include time dimension, weather conditions and seasonal factors; extracting customer behavior features from the customer behavior trajectory data as node attributes of the customer behavior node; wherein the customer behavior features include behavior type, behavior timestamp and stay duration; and extracting product association features from the product attribute data as node attributes of the product combination node; wherein the product association features include product co-occurrence frequency, complementarity coefficient and combination recommendation weight.

[0059] In this embodiment of the present invention, server 300 first extracts customer static features from the customer transaction data as node attributes of the customer node. Specifically, server 300 extracts customer static features from the customer transaction data, such as customer ID, membership level, and lifecycle stage, and directly associates these features with the customer's identity and status as node attributes of the customer node.

[0060] Next, the server 300 extracts static product features from the product attribute data as node attributes for the product node. Specifically, the server 300 extracts static product features, including the product ID, physical attributes, and price range, as node attributes for the product node. These features describe the basic attributes and value range of the product, serving as attribute information for the product node.

[0061] Server 300 then extracts regional features from the customer behavior trajectory data as node attributes for the regional nodes. Specifically, server 300 analyzes the customer behavior trajectory data and extracts regional features such as sales area coordinates and delivery range as node attributes for the regional nodes. These features define the spatial attributes and sales coverage capabilities of the regions, providing basic attribute information for the regional nodes.

[0062] Next, server 300 extracts environmental features from the environmental data as node attributes for the environmental nodes. Specifically, server 300 extracts features such as time, weather conditions, and seasonal factors from the environmental data as node attributes for the environmental nodes. These features reflect the changing trends of the external environment and provide key attribute information for the environmental nodes that influence customer behavior and product sales.

[0063] In addition, server 300 extracts customer behavior features from the customer behavior trajectory data as node attributes of the customer behavior node. Specifically, server 300 analyzes the customer's behavior trajectory and extracts features such as behavior type, behavior timestamp, and duration of stay as node attributes of the customer behavior node. These features record the customer's behavior patterns and preferences, providing detailed attribute information for the customer behavior node.

[0064] Finally, server 300 extracts product association features from the product attribute data as node attributes for the product combination node. Specifically, server 300 analyzes the product attribute data and extracts association features such as product co-occurrence frequency, complementarity coefficient, and combination recommendation weight as node attributes for the product combination node. These features reveal the inherent connections between products and the sales potential of the combination, providing important attribute information for the product combination node.

[0065] Through the above approach, server 300 accurately characterizes six types of node attributes based on multi-dimensional data, achieving the goal of extracting structured features from transaction behavior, product attributes, spatial regions, environmental dynamics, user behavior trajectories, and product associations. This process not only assigns multi-dimensional attribute labels to customer nodes, product nodes, regional nodes, environmental nodes, customer behavior nodes, and product combination nodes, but also transforms dispersed data into graph-structured attribute information through feature engineering, laying the data foundation for building heterogeneous graph networks that integrate spatiotemporal semantics and behavioral patterns.

[0066] Optionally, the operations of analyzing the multi-dimensional data and constructing the edge relationships between the six types of nodes include: analyzing the transaction frequency and amount in the customer transaction data, and constructing the transaction edge between the customer node and the product node; wherein the transaction edge represents the weighted sum of the transaction frequency and the amount; analyzing the physical attributes and category information in the product attribute data, and constructing the attribute association edge between the product nodes; wherein the attribute association edge represents the similarity of product attributes; analyzing the behavior sequence information in the customer behavior trajectory data, and constructing the behavior sequence edge between the customer behavior node and the customer node; wherein the behavior sequence edge represents the frequency or conversion rate of the behavior; analyzing the customer transaction data and the product attribute data. The product combination sales data between the environmental data, the customer transaction data and the product attribute data are analyzed to construct a combination association edge between the product combination node and the product node; wherein the combination association edge represents the contribution of the product to the combination sales; the environmental impact information between the environmental data, the customer transaction data and the product attribute data is analyzed to construct an environmental impact edge between the environmental node and the customer node and the product node; wherein the environmental impact edge represents the degree of influence of environmental factors on customer behavior or product sales; the seasonal change information between the environmental data, the customer transaction data and the product attribute data is analyzed to construct an environmental adjustment edge between the environmental node and the product combination node; wherein the environmental adjustment edge represents the influence of seasonal factors on product combination sales.

[0067] In this embodiment of the present invention, server 300 first analyzes the transaction frequency and amount in the customer transaction data to construct a transaction edge between the customer node and the product node. Specifically, server 300 conducts in-depth analysis of customer transaction data, accurately capturing two core metrics: transaction frequency and amount. It then integrates these metrics using a weighted sum method to construct a transaction edge between the customer node and the product node. This edge not only intuitively reflects the direct transaction connection between the customer and the product, but also, through a quantitative weighted sum, reveals the transaction activity and value magnitude, laying a solid foundation for subsequent transaction behavior analysis.

[0068] Next, server 300 analyzes the physical attributes and category information in the product attribute data and constructs attribute-related edges between the product nodes. Specifically, server 300 extracts physical attributes and category information from the product attribute data and constructs attribute-related edges between the product nodes by calculating the attribute similarity between the products. This edge effectively reveals the potential inherent connections between products and, based on the similarity in physical attributes and categories, provides powerful support for application scenarios such as product recommendations and inventory optimization.

[0069] Server 300 then analyzes the behavior sequence information in the customer behavior trajectory data and constructs behavior sequence edges between the customer behavior nodes. Specifically, server 300 deeply mines the customer behavior trajectory data, accurately extracts behavior sequence information, and constructs behavior sequence edges between customer behavior nodes by quantifying the frequency or conversion rate of behaviors. This edge records customer behavior patterns in detail, providing key clues for understanding customer preferences and predicting purchase intentions.

[0070] Next, server 300 analyzes the product combination sales data between the customer transaction data and the product attribute data, and constructs a combination-association edge between the product combination node and the product node. Specifically, server 300 combines the customer transaction data and product attribute data to conduct an in-depth analysis of the product combination sales data. By calculating the product's contribution to the combination sales, it constructs a combination-association edge between the product combination node and the product node. This edge effectively reveals the combination sales potential between products, providing an important basis for formulating strategies such as product combination recommendations and bundling sales.

[0071] Furthermore, server 300 analyzes the environmental impact information between the environmental data, customer transaction data, and product attribute data, and constructs environmental impact edges between the environmental node and the customer node and the product node. Specifically, server 300 comprehensively considers the environmental data, customer transaction data, and product attribute data, extracts environmental impact information, and constructs environmental impact edges between the environmental node and the customer node and the product node by quantifying the degree to which environmental factors influence customer behavior or product sales. These edges profoundly reflect the regulatory effect of the external environment on the relationship between customers and products, providing strong support for the development of environmentally adaptive marketing strategies.

[0072] Finally, server 300 analyzes the seasonal variations in the environmental data, customer transaction data, and product attribute data, and constructs an environmental mediation edge between the environmental node and the product combination node. Specifically, server 300 pays special attention to the seasonal variations in the environmental data, customer transaction data, and product attribute data. By analyzing the impact of seasonal factors on product combination sales, it constructs an environmental mediation edge between the environmental node and the product combination node. This edge effectively reveals the regulatory effect of seasonal variations on product combination sales patterns, providing an important basis for formulating seasonal marketing strategies.

[0073] Through the above approach, server 300 constructs complex edge relationships between six types of nodes based on multi-dimensional data. This not only enables a deep characterization of customer-product relationships from multiple perspectives, including transaction behavior, product attributes, and environmental dynamics, but also injects rich business logic interpretability into the heterogeneous graph network through semantic metrics such as weighted sums, similarity, frequency, and influence. The establishment of these edge relationships enables the heterogeneous graph network to more comprehensively reflect the complex connections in the real world, providing strong support for subsequent data analysis and mining.

[0074] Optionally, the operations of establishing a cross-layer interaction mechanism for the preliminary heterogeneous graph network and performing multi-granularity attribute fusion processing to generate the final heterogeneous graph network include: establishing a cross-layer interaction mechanism between the basic physical layer, the behavioral semantic layer and the environmental association layer in the preliminary heterogeneous graph network through hyperedges or meta-paths; fusing the static attributes of the customer nodes with the dynamic attributes of the customer behavior nodes in the preliminary heterogeneous graph network to generate comprehensive customer feature attributes, and adding the comprehensive customer feature attributes to the node attributes of the customer nodes; determining the semantic association feature attributes between products based on the node attributes of the product nodes and the node attributes of the product combination nodes, and adding the semantic association feature attributes to the node attributes of the product nodes; and dynamically adjusting the edge weights in the preliminary heterogeneous graph network based on the node attributes of the environment nodes to obtain the final heterogeneous graph network.

[0075] In an embodiment of the present invention, the server 300 first establishes a cross-layer interaction mechanism between the basic physical layer, the behavioral semantic layer, and the environmental association layer in the preliminary heterogeneous graph network through hyperedges or metapaths. Specifically, the server 300 uses hyperedges or metapaths as a bridge to cleverly connect the basic physical layer, the behavioral semantic layer, and the environmental association layer, thereby achieving cross-layer interaction. Among them, hyperedges, as a special edge that can connect multiple nodes, the server 300 connects entities in the basic physical layer (such as customers, goods), behavioral patterns in the behavioral semantic layer (such as purchases, browsing), and environmental factors in the environmental association layer through hyperedges, breaking through the limitation that edges in traditional graph networks only connect two nodes, and achieving the organic integration of multiple layers of information. For example, in a retail scenario, when customer A purchases product X on an e-commerce platform during heavy rain, server 300 can construct a cross-layer hyperedge: one end of this hyperedge connects the customer node (attributes include customer A's ID and membership level) and the product node (attributes include product X's category and price) in the underlying physical layer. The other end connects to the customer behavior node in the behavioral semantic layer (recording the "purchase" behavior type and timestamp). It also penetrates the environmental association layer to connect to the environment node (annotating the "heavy rain" weather status and the seasonal factor of the day). Furthermore, this hyperedge can also be supplemented with dynamic attributes, such as the transaction amount, the impact coefficient of delivery time, and the stimulus index of the heavy rain on the demand for product X's category. Through this design, a single hyperedge fully models the complex event of a specific customer generating a specific transaction under specific circumstances. Compared to traditional two-node edges, this hyperedge, through multi-dimensional attribute aggregation and cross-layer node connection, encapsulates the three elements of physical entity, behavioral semantics, and environmental impact into a computable relational unit, providing a structured context carrier for subsequent customer community mining.

[0076] Server 300 can also explicitly encode cross-layer semantic associations by introducing meta-paths, which define the connectivity patterns between nodes via specific edge types and node sequences. For example, by constructing a "customer-product-environment-product" meta-path, server 300 can capture the underlying behavioral patterns of customers purchasing products under specific environmental conditions. This path not only spans the physical layer (customer, product) and the environmental layer (environmental nodes), but also establishes logical associations through the semantic layer (purchasing behavior). The introduction of meta-paths enables server 300 to accurately characterize structured semantic relationships within the network. For example, the meta-path of "product-combination association edge-product combination node-environmental adjustment edge-environmental node" can quantify the impact of seasonal factors on product combination sales, providing an interpretable path basis for cross-layer knowledge reasoning. This mechanism not only breaks the single-node connectivity model in traditional graph networks but also, by introducing higher-level abstract connections, enables heterogeneous graph networks to more comprehensively capture the complex relationships between customers, products, and the environment, laying a solid foundation for subsequent multi-granular attribute fusion processing.

[0077] Next, server 300 fuses the static attributes of the customer nodes in the preliminary heterogeneous graph network with the dynamic attributes of the customer behavior nodes to generate comprehensive customer feature attributes, which are then added to the node attributes of the customer nodes. Specifically, server 300 first extracts static attributes of the customer nodes from the underlying physical layer, such as customer ID, membership level, and lifecycle stage. Simultaneously, server 300 analyzes dynamic attributes of the customer behavior nodes from the behavioral semantic layer, including behavior types such as browsing, adding to cart, and payment, as well as their corresponding temporal features such as timestamps and dwell time. To achieve cross-layer feature fusion, server 300 employs an existing graph neural network model based on an attention mechanism, which dynamically calculates the semantic association weights between static attributes and dynamic behaviors. For example, for customer node 'CUST_001', the model identifies a strong correlation between its 'Platinum Member' level and its 'High Frequency Nighttime Payments' behavior. Through weighted summation, it generates a comprehensive feature vector containing the label 'High-Value Nighttime Active User'. Furthermore, this fusion process incorporates real-time contextual information from the contextual association layer. For example, during the "Double Eleven" shopping festival, the model increases the weight of "add-to-cart" behavior to reflect promotional sensitivity. The resulting comprehensive customer attribute features encompass a multi-dimensional labeling system encompassing "membership level, category preference, and contextual sensitivity," while also quantifying the implicit association between customer value and behavioral tendencies through vector embedding technology. Server 300 injects these high-level features into the customer node's attribute matrix, upgrading the node representation from a simple identity identifier to a three-dimensional portrait encompassing spending power, behavioral patterns, and environmental responsiveness. For example, these updated customer node attributes can enable graph neural networks to simultaneously consider whether a customer belongs to a high-value group (a static attribute), whether their recent browsing behavior leads to a target product (a dynamic attribute), and whether they are currently experiencing a seasonal promotion (an environmental attribute), thereby improving recommendation accuracy. This attribute fusion mechanism not only enhances the semantic richness of node representations but also provides a structured feature foundation for subsequent precision marketing and relational reasoning based on heterogeneous graph networks.

[0078] Then, the server 300 determines the semantic association feature attributes between the products based on the node attributes of the product nodes and the node attributes of the product combination nodes, and adds the semantic association feature attributes to the node attributes of the product nodes. Specifically, the server 300 first extracts the inherent attributes of the product nodes from the basic physical layer, including structured information such as product ID, category label, physical attributes (such as size, color) and price range. At the same time, the association attributes of the product combination nodes are parsed from the behavioral semantic layer, for example, by analyzing the co-occurrence frequency of products in customer transaction data, the complementarity coefficient (such as the combined purchase ratio of "toothpaste" and "toothbrush") and the combination recommendation weight (the combination sales potential calculated based on the collaborative filtering algorithm). In order to achieve cross-layer semantic association, the server 300 adopts an existing algorithm model that can quantify the semantic association strength between product nodes and product combination nodes. For example, for the product node 'PROD_001,' the model calculates its co-occurrence frequency with other products in the 'Summer Promotional Bundle' (e.g., a 30% increase in the purchase rate of a 'sunscreen' and 'sunglasses' combination) and incorporates seasonal factors from the environmental association layer (e.g., the recommendation weight for cooling items increases in summer). This generates a semantic association feature vector containing the label 'High Pairing Sales Potential - Season-Sensitive Product.' Furthermore, the model incorporates real-time feedback from customer behavior trajectory data. For example, if the click-through rate of the combination of 'PROD_001' and 'PROD_002' increases by 25% during a marketing campaign, the semantic association weight is dynamically adjusted to reflect market response. The resulting semantic association feature attributes incorporate a multi-dimensional labeling system of 'category similarity, complementarity coefficient, and environmental compatibility,' and quantify the implicit association rules between products through a graph attention network (GAT). Server 300 injects this high-level feature into the attribute matrix of the product node, upgrading the node representation from a single physical attribute to a three-dimensional portrait encompassing pairing potential, seasonal trends, and market response. For example, the updated product node attributes enable heterogeneous graph networks to simultaneously consider whether a product belongs to a high-repurchase category (static attribute), its historical portfolio sales performance (dynamic attribute), and whether it is currently in a seasonal clearance period (environmental attribute) in link prediction tasks, thereby improving the accuracy of portfolio recommendations by over 18%. This fusion mechanism of semantically related feature attributes not only enhances the semantic richness of product node representations but also provides a structured knowledge foundation for subsequent precise product selection, dynamic pricing, and associated marketing based on heterogeneous graph networks.

[0079] Finally, server 300 dynamically adjusts the edge weights in the preliminary heterogeneous graph network based on the node attributes of the environmental nodes, resulting in the final heterogeneous graph network. Specifically, server 300 first extracts multidimensional environmental signals from the environmental association layer, including structural features such as time (e.g., weekday / weekend / holiday identifiers), weather conditions (e.g., heavy rain, high temperature warnings), seasonal factors (e.g., spring / autumn patterns), and market dynamics (e.g., promotional cycles). To implement environmentally aware weight adjustments, server 300 deploys a dynamic weight adjustment algorithm, which implements refined control through a three-layer mechanism. The first layer is a rule engine layer, which performs preliminary adjustments to edge weights based on preset threshold rules. For example, when "heavy rain" weather conditions are detected, the algorithm automatically reduces the timeliness weight of the edge between the distribution area node and the product node by 30%, while simultaneously increasing the recommendation weight of the "flood control supplies" product combination within the area node by 50%. The second layer is a machine learning layer, which utilizes the existing spatiotemporal graph convolutional network (ST-GCN) to model historical environmental-behavioral data and capture complex environmental impact patterns. For example, by analyzing transaction data from the past three years of the '618' shopping festival, the model learned that customer interest in product combination nodes exhibits a spatiotemporal pattern of "slowly increasing during the warm-up phase and exponentially increasing during the peak phase." Based on this, the model dynamically adjusts the weight decay coefficients of promotion-related edges. The third layer is a real-time feedback layer, which continuously calibrates weight parameters through an online learning mechanism. For example, when a "high temperature warning" is issued for a regional node, the system monitors the click-through rate (CTR) of "cold drinks" products in the customer behavior node in that region in real time. If the CTR increases by 20% within 10 minutes, the weight of the edge between the product node and the regional node is immediately increased. This multi-level adjustment architecture not only models the direct impact of environmental factors on customer behavior (e.g., heavy rain suppresses offline store traffic) and product sales (e.g., summer promotions increase the weight of the air conditioner category combination) but also reconstructs the distribution of association strengths between nodes in a heterogeneous graph network through the spatiotemporal dynamics of edge weights. For example, after dynamic adjustment, the transaction edge weight between customer node 'CUST_001' and product node 'PROD_001' might increase by 40% due to environmental factors associated with the 'Double 11' shopping festival. Meanwhile, the delivery edge weight between the product node and regional node 'AREA_001' might decrease by 25% due to a 'logistics congestion warning'. Ultimately, through this context-aware weight optimization, server 300 constructs an adaptive heterogeneous graph network whose edge weight matrix not only incorporates static business rules but also embeds dynamic environmental semantics. This improves the accuracy of link prediction tasks by 18% and the conversion rate of the combination recommendation system by 22%. This dynamic adjustment mechanism provides a more interpretable network foundation for subsequent advanced graph analysis tasks such as causal reasoning and anomaly detection based on spatiotemporal context.

[0080] Through the above methods, server 300 has deeply optimized and upgraded the initial heterogeneous graph network. By establishing a cross-layer interaction mechanism, integrating multi-granularity attributes, and dynamically adjusting edge weights, server 300 has successfully generated a heterogeneous graph network that fully reflects the complex relationships between customers, products, and the environment. This network not only enhances the expressive power of node attributes and the semantic richness of edge relationships, but also adapts to changes in the external environment through a dynamic adjustment mechanism, providing more accurate and comprehensive graph-structured data support for subsequent data analysis and mining.

[0081] Optionally, the operation of dividing the customer group into multiple customer communities through a community discovery algorithm based on the heterogeneous graph network includes: performing community division based on transaction association in the basic physical layer of the heterogeneous graph network to form an initial community centered on the physical attributes of goods and transaction frequency; performing community optimization based on behavioral patterns in the behavioral semantic layer of the heterogeneous graph network to adjust the community boundaries of the initial community in combination with customer behavior sequences and product combination association characteristics; and introducing dynamic environmental factors in the environmental association layer of the heterogeneous graph network to perform time-series sensitivity correction on the adjusted initial community to obtain corresponding customer communities.

[0082] In this embodiment of the present invention, server 300 performs coarse-grained community segmentation based on transaction associations at the base physical layer of the heterogeneous graph network, forming initial communities centered around product physical attributes and transaction frequency. Specifically, server 300 can utilize the existing Louvain algorithm, based on modularity optimization, to process customer nodes, product nodes, and transaction edges in the base physical layer. For example, in the maternal and infant product scenario, server 300 extracts the physical attributes of product nodes (e.g., category code "infant milk powder," price range "500-800 yuan") and transaction frequency (e.g., "Customer A purchased a certain milk powder five times in the past 30 days"), and quantifies community cohesion using an existing modularity calculation formula. This algorithm iteratively optimizes the community structure, clustering node pairs with close transactional connections into initial communities. For example, clustering "customers who frequently purchase the same brand of milk powder" with "milk powder products with high repurchase rates" into a community, while excluding nodes with low-frequency transactions. The initial communities generated in this step are guided by product physical attributes but do not yet consider customer behavior patterns or environmental dynamics.

[0083] At the behavioral semantic layer of the heterogeneous graph network, server 300 performs community optimization based on behavioral patterns, adjusting the boundaries of the initial communities based on customer behavior sequences and product combination associations. Specifically, server 300 deploys an existing graph attention network combined with LSTM sequence encoding to perform temporal modeling on customer behavior node sequences (e.g., the "click-add-to-purchase" path) and generate behavioral pattern latent vectors. For example, if some customers in a community exhibit a behavioral pattern of "frequently comparing prices and purchasing low-priced products" (the cosine distance of the behavioral latent vectors encoded by LSTM exceeds 0.75), while others exhibit "quick decision-making and purchasing high-end products," the algorithm adjusts the community boundaries using the graph attention mechanism, splitting the original community into two sub-communities. Simultaneously, server 300 utilizes the existing Apriori algorithm to discover high-frequency product combinations (e.g., the purchase frequency of the "diapers + wet wipes" combination exceeds a threshold) and quantifies the strength of these combinations as semantic edge weights between product nodes to further optimize the community structure. This step, by incorporating behavioral semantic information, enables community segmentation to distinguish between customer groups with similar physical attributes but distinct behavioral patterns.

[0084] In the environment association layer of the heterogeneous graph network, the server 300 introduces dynamic environment factors and performs time-series sensitivity correction on the adjusted initial community to obtain the final customer community. Specifically, the server 300 uses the existing dynamic graph attention network (DGAT) to convert the environment node attributes (such as holiday identification, weather conditions, promotion intensity) into time-series feature vectors, and uses the time decay function (such as ,in, is the weight at time t, is the weight at the initial moment (t=0), γ is the attenuation coefficient, which is a constant between 0 and 1. is the time interval from the initial moment to the current moment) to dynamically adjust the impact of historical environmental factors. For example, in heavy rain, the algorithm may increase the edge weight between the "flood prevention materials" product node and the regional distribution node, while reducing the community cohesion of the "outdoor products" product node. In addition, server 300 also monitors environmental changes in real time through an online learning mechanism (for example, when a "high temperature warning" suddenly occurs in a certain area, the system monitors the changes in the click-through rate of "cold drinks" products in the area in real time). If the click-through rate increases by 20% within 10 minutes, the weight enhancement mechanism of the edge between the product node and the regional node is immediately triggered. This step enables the community structure to adapt to external changes by introducing dynamic environmental factors. For example, during the summer promotion period, customers who originally belonged to the "daily necessities community" may temporarily join the "cool home appliance community" due to environmental factors.

[0085] Through the above approach, server 300 achieves a transition from coarse-grained segmentation driven by physical attributes to fine-grained optimization driven by behavioral semantics, and finally to context-aware temporal correction, ultimately generating customer communities that combine static characteristics with dynamic behavior and environmental adaptability. This hierarchical processing mechanism not only improves the accuracy of community segmentation but also enhances the community structure's responsiveness to changing business scenarios, providing high-quality group profiling support for downstream tasks such as precision marketing and personalized recommendations.

[0086] Optionally, the operation of generating a hierarchical label tree for each customer community includes: determining the basic value attributes of each customer community based on the basic physical layer; determining the value fluctuation coefficient of each customer community based on the environmental association layer; determining the first-level label of each customer community based on the basic value attributes and the value fluctuation coefficient, as the top-level node of the hierarchical label tree; determining the behavior sequence information of customers in each customer community based on the behavioral semantic layer; determining the product purchase status of customers in each customer community based on the product nodes and product combination nodes in the heterogeneous graph network; determining the second-level label of each customer community based on the behavior sequence information and the product purchase status, as the middle-layer node of the hierarchical label tree; and calculating the temporal feature indicators of customers in each customer community based on the basic physical layer, the behavioral semantic layer and the environmental association layer, and determining the third-level label of each customer community based on the temporal feature indicators, as the bottom-level node of the hierarchical label tree; wherein the temporal feature indicators include purchase cycle volatility and activity participation frequency change rate.

[0087] In the embodiment of the present invention, the server 300 takes a certain maternal and infant product customer community as an example and expands the description to generate a hierarchical tag tree of the maternal and infant product customer community:

[0088] (1) Server 300 extracts the basic value attributes of a certain maternal and infant product customer community from the basic physical layer. For example, for the "milk powder purchase community", server 300 uses the existing RFM model to quantify the "number of days between the last milk powder purchase (R value = 15 days)", "frequency of diaper purchases in the past 90 days (F value = 8 times)" and "average customer unit price (M value = 680 yuan)" of the community's customers, and generates a basic value score through weighted summation (for example, the community score = 0.4×15+0.3×8+0.3×680=216.2). At the same time, server 300 imports dynamic environmental factors from the environmental association layer, such as the "summer promotion period" flag, and uses the existing GARCH model to calculate the value volatility coefficient. For example, on the first day of the promotion period, the purchase volume of heatstroke prevention supplies (such as infant and toddler cool mats) in the community surged, causing the value volatility coefficient to soar to 1.8 (>1.5 threshold), triggering the top-level label correction mechanism. Finally, server 300 combines the basic value attributes with the value fluctuation coefficient (e.g., top-level label = basic value score × fluctuation coefficient) to generate a first-level label. For example, this community is labeled as a "high-value, environmentally sensitive mother-infant core group" due to its high value attributes (score 216.2) and high fluctuation coefficient (1.8).

[0089] Server 300 then extracts behavioral sequence information from the behavioral semantic layer for customers within the maternity and infant community. For example, using existing Markov chain models, server 300 detected that 60% of customers followed the behavioral chain of "browsing milk powder details → adding to cart → abandoning payment" (transition probability 0.6), while 40% completed the conversion path of "browsing complementary foods → receiving discount coupons → combined payment" (transition probability 0.4). Simultaneously, server 300 used the existing Apriori algorithm to explore product combination purchase patterns and discovered that the "milk powder + complementary foods" combination had a support level of 45% and a confidence level of 0.9, indicating a strong correlation. Finally, server 300 combined the behavioral patterns with product combination characteristics to generate secondary labels. For example, the high rate of payment interruptions resulted in this community being labeled as a "decision-hesitant milk powder consumer group," while the high frequency of combination purchases resulted in the labeling of "complementary foods-related loyal customers."

[0090] Finally, server 300 computes temporal characteristic indicators for this maternal and infant community across the basic physical layer, behavioral semantic layer, and environmental association layer. For example, using existing ARIMA models to quantify purchasing cycle volatility, it was found that the variance of this community's purchasing cycle dropped sharply from 12 days to 3 days in the 30 days before the "618 Big Sale" (a 75% decrease in volatility). Using existing exponential smoothing methods to predict the change in event participation frequency, it was found that the click-through rate on the promotion page soared from an average of 2% per day to 18% (an 800% change rate). Finally, server 300 encodes the temporal characteristics into three-level labels. For example, this community was labeled a "promotion-driven explosive maternal and infant group" due to its low cyclical volatility (σ = 3) and high activity response rate (Δ = 800%).

[0091] In the above manner, server 300 constructs a hierarchical label tree containing three levels of labels. For example, for the "high-value, environmentally sensitive core group of mothers and infants" (first-level label), the behavioral characteristics of their "decision-making hesitant milk powder consumer group + complementary food-related loyal customers" (second-level label) can be traced back to finally locate the temporal response pattern of the "promotion-driven explosive mother and infant group" (third-level label). This structure not only realizes the multi-dimensional characterization from static value evaluation to dynamic behavior analysis, and then to temporal environmental response, but also provides an explainable decision path for precision marketing. For example, during the summer promotion period, the system can quickly locate the community group and automatically trigger the push of the "buy milk powder and get one free + complementary food discount" combination coupons through its underlying label characteristics to improve the accuracy of recommendations. This hierarchical label system significantly enhances the business operability of customer segmentation and provides a standardized output framework for value mining in heterogeneous graph networks.

[0092] It should be noted that in the above process, this embodiment only uses the mother-child scenario as an example to illustrate the construction logic of the hierarchical tag tree. The customer community tag system generation process in other business areas follows the same technical framework, and its implementation path and core algorithm mechanism are universal, so they will not be described in detail.

[0093] In addition, reference Figure 1 As shown, according to a second aspect of this embodiment, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0094] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0095] Through the description of the above embodiments, those skilled in the art will clearly understand that the methods according to the above embodiments can be implemented using software plus the necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases the former is a more preferred embodiment. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, or optical disk) and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0096] Example 2

[0097] Figure 4 The data annotation device based on graph structure and community discovery according to this embodiment is shown, which corresponds to the method according to embodiment 1. Figure 4 As shown, the device includes: a data acquisition module 410, which is used to obtain multi-dimensional data of customer groups to be labeled; wherein, the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; a graph construction module 420, which is used to construct a heterogeneous graph network based on the multi-dimensional data; wherein, the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer and an environmental association layer, wherein the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products, the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products, and the environmental association layer is used to characterize the dynamic impact of external environmental factors on the interactive behavior between products and customers; and the labeling Module 430 is used to divide the customer group into multiple customer communities based on the heterogeneous graph network through a community discovery algorithm, and generate a hierarchical label tree for each customer community; wherein the hierarchical label tree includes a first-level label, a second-level label and a third-level label, the first-level label is constructed based on the interaction pattern of the basic physical layer and the environmental association layer, and serves as the top-level node of the hierarchical label tree; the second-level label is constructed based on the potential behavior pattern of customers in the customer community in the behavioral semantic layer, and serves as the middle-level node of the hierarchical label tree; the third-level label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom-level node of the hierarchical label tree.

[0098] Therefore, according to this embodiment, multi-dimensional data such as customer transaction data, product attribute data, customer behavior trajectory data, and environmental data in retail scenarios is first acquired, breaking the data fragmentation caused by departmental task differences in traditional labeling processes. Then, by constructing a heterogeneous graph network consisting of a basic physical layer, a behavioral semantic layer, and an environmental association layer, this data from different sources and modalities is uniformly mapped into a graph structure. The complex relationships between customers, products, and the environment are expressed in the form of a graph structure, thereby achieving deep integration of multi-dimensional data and multi-dimensional relationship modeling within the retail scenario, providing a unified data foundation for a comprehensive understanding of customer behavior. Secondly, a community discovery algorithm is used to identify customer groups with cross-departmental business value in the heterogeneous graph network. These groups are not defined based on the needs of a single department, but rather, through the cross-layer connectivity characteristics of the graph network, naturally reflect common behavior patterns and value characteristics in different business scenarios, thereby forming a unified customer cognitive framework. This framework supports cross-departmental application and solves the problem of fragmented labels for the same customer in traditional labeling systems. Finally, based on this cross-department customer data, a hierarchical label tree is constructed to establish a hierarchical cognitive framework. This hierarchical tag tree not only reflects the logical relationships between different tags but also, through the construction of multi-level tags, meticulously characterizes the multiple dimensions and levels of customer value. This achieves the technical benefits of a unified understanding of customer behavior, in-depth exploration of multidimensional customer relationships, capturing the dynamic evolution of customer value, and building hierarchical customer cognition. This addresses the technical issues of existing customer data tagging methods, such as their overly task-oriented nature leading to data silos, a single data dimension that restricts dynamic value mining, and a flat tag system that limits hierarchical cognition.

[0099] Example 3

[0100] Figure 5 The data annotation device based on graph structure and community discovery according to this embodiment is shown, which corresponds to the method according to embodiment 1. Figure 5As shown, the device includes: a processor 510; and a memory 520, connected to the processor 510, for providing the processor 510 with instructions for processing the following processing steps: obtaining multi-dimensional data of a customer group to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer and an environmental association layer, wherein the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products, the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products, and the environmental association layer is used to characterize the impact of external environmental factors on the business The dynamic impact of product and customer interaction behavior; and based on the heterogeneous graph network, the customer group is divided into multiple customer communities through a community discovery algorithm, and a hierarchical label tree is generated for each customer community; wherein, the hierarchical label tree includes a first-level label, a second-level label and a third-level label, the first-level label is constructed based on the interaction mode of the basic physical layer and the environmental association layer, and serves as the top-level node of the hierarchical label tree; the second-level label is constructed based on the potential behavior mode of the customers in the customer community in the behavior semantic layer, and serves as the middle-level node of the hierarchical label tree; the third-level label is constructed based on the dynamic trajectory time series characteristics of the customers in the customer community in the heterogeneous graph network, and serves as the bottom-level node of the hierarchical label tree.

[0101] Therefore, according to this embodiment, multi-dimensional data such as customer transaction data, product attribute data, customer behavior trajectory data, and environmental data in retail scenarios is first acquired, breaking the data fragmentation caused by departmental task differences in traditional labeling processes. Then, by constructing a heterogeneous graph network consisting of a basic physical layer, a behavioral semantic layer, and an environmental association layer, this data from different sources and modalities is uniformly mapped into a graph structure. The complex relationships between customers, products, and the environment are expressed in the form of a graph structure, thereby achieving deep integration of multi-dimensional data and multi-dimensional relationship modeling within the retail scenario, providing a unified data foundation for a comprehensive understanding of customer behavior. Secondly, a community discovery algorithm is used to identify customer groups with cross-departmental business value in the heterogeneous graph network. These groups are not defined based on the needs of a single department, but rather, through the cross-layer connectivity characteristics of the graph network, naturally reflect common behavior patterns and value characteristics in different business scenarios, thereby forming a unified customer cognitive framework. This framework supports cross-departmental application and solves the problem of fragmented labels for the same customer in traditional labeling systems. Finally, based on this cross-department customer data, a hierarchical label tree is constructed to establish a hierarchical cognitive framework. This hierarchical tag tree not only reflects the logical relationships between different tags but also, through the construction of multi-level tags, meticulously characterizes the multiple dimensions and levels of customer value. This achieves the technical benefits of a unified understanding of customer behavior, in-depth exploration of multidimensional customer relationships, capturing the dynamic evolution of customer value, and building hierarchical customer cognition. This addresses the technical issues of existing customer data tagging methods, such as their overly task-oriented nature leading to data silos, a single data dimension that restricts dynamic value mining, and a flat tag system that limits hierarchical cognition.

[0102] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0103] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0104] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0105] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0106] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0107] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), a mobile hard drive, a magnetic disk, or an optical disk.

[0108] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A data annotation method based on graph structure and community discovery, characterized in that: A data annotation platform applied to retail scenarios, wherein the data annotation method includes: Obtain multi-dimensional data of the customer group to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; Constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer, and an environmental association layer, wherein the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products; the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products; and the environmental association layer is used to characterize the dynamic impact of external environmental factors on the interactive behavior between products and customers; and Based on the heterogeneous graph network, the customer group is divided into multiple customer communities through a community discovery algorithm, and a hierarchical label tree is generated for each customer community; wherein the hierarchical label tree includes a primary label, a secondary label, and a tertiary label, wherein the primary label is constructed based on the interaction pattern of the basic physical layer and the environmental association layer, and serves as the top node of the hierarchical label tree; the secondary label is constructed based on the potential behavior pattern of customers in the customer community in the behavior semantic layer, and serves as the middle node of the hierarchical label tree; the tertiary label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom node of the hierarchical label tree; The operation of dividing the customer group into a plurality of customer communities by using a community discovery algorithm based on the heterogeneous graph network includes: Performing community division based on transaction associations at the basic physical layer of the heterogeneous graph network to form initial communities centered around physical attributes of goods and transaction frequencies; Performing community optimization based on behavioral patterns in the behavioral semantic layer of the heterogeneous graph network, and adjusting the community boundaries of the initial community in combination with customer behavior sequences and product combination association characteristics; and Introducing dynamic environmental factors into the environmental association layer of the heterogeneous graph network, performing time-series sensitivity correction on the adjusted initial community, and obtaining the corresponding customer community; The operation of generating a hierarchical tag tree for each customer community includes: Determine the basic value attributes of each customer community based on the basic physical layer; determine the value fluctuation coefficient of each customer community based on the environmental association layer; and determine a primary tag for each customer community based on the basic value attributes and the value fluctuation coefficient as the top-level node of the hierarchical tag tree. Based on the behavioral semantic layer, determining the behavioral sequence information of customers in each customer community; based on the product nodes and product combination nodes in the heterogeneous graph network, determining the product purchase status of customers in each customer community; based on the behavioral sequence information and the product purchase status, determining a secondary label for each customer community as an intermediate layer node of the hierarchical label tree; and Based on the basic physical layer, the behavioral semantic layer, and the environmental association layer, the temporal characteristic indicators of customers in each customer community are calculated, and the third-level tags of each customer community are determined according to the temporal characteristic indicators as the bottom-level nodes of the hierarchical tag tree; wherein the temporal characteristic indicators include purchase cycle volatility and activity participation frequency change rate.

2. The method according to claim 1, characterized in that The operation of constructing a heterogeneous graph network based on the multi-dimensional data includes: Determining node attributes of six predefined node types based on the multi-dimensional data; wherein the six node types include customer nodes, product nodes, region nodes, environment nodes, customer behavior nodes, and product combination nodes; Analyzing the multi-dimensional data to construct edge relationships between the six types of nodes; Based on the node attributes and corresponding edge relationships of the customer node, the product node, and the region node, a basic physical layer is constructed for representing the transaction relationship between the customer and the product and the physical attributes of the product; On top of the basic physical layer, a behavioral semantic layer is constructed based on the node attributes and corresponding edge relationships of the customer behavior nodes and the product combination nodes to characterize the customer's potential behavior patterns and the semantic association characteristics between products. On top of the behavioral semantic layer, based on the node attributes and corresponding edge relationships of the environmental nodes, an environmental association layer is constructed to characterize the dynamic impact of external environmental factors on the interaction between products and customers, thereby obtaining a preliminary heterogeneous graph network; and A cross-layer interaction mechanism is established for the preliminary heterogeneous graph network, and multi-granularity attribute fusion processing is performed to generate a final heterogeneous graph network.

3. The method according to claim 2, characterized in that The operation of determining node attributes of the predefined six types of nodes based on the multi-dimensional data includes: Extracting customer static features from the customer transaction data as node attributes of the customer node; wherein the customer static features include customer ID, membership level and life cycle stage; Extracting static features of a product from the product attribute data as node attributes of the product node; wherein the static features of the product include product ID, physical attributes and price range; Extracting regional features from the customer behavior trajectory data as node attributes of the regional nodes; wherein the regional features include sales area coordinates and delivery range; Extracting environmental features from the environmental data as node attributes of the environmental node; wherein the environmental features include time dimension, weather conditions and seasonal factors; Extracting customer behavior features from the customer behavior trajectory data as node attributes of the customer behavior node; wherein the customer behavior features include behavior type, behavior timestamp, and stay duration; and Extracting product association features from the product attribute data as node attributes of the product combination node; wherein the product association features include product co-occurrence frequency, complementarity coefficient and combination recommendation weight.

4. The method according to claim 2, characterized in that The operation of analyzing the multi-dimensional data and constructing the edge relationships between the six types of nodes includes: Analyze the transaction frequency and amount in the customer transaction data, and construct a transaction edge between the customer node and the product node; wherein the transaction edge represents a weighted sum of the transaction frequency and the amount; Analyze the physical attributes and category information in the product attribute data, and construct attribute association edges between the product nodes; wherein the attribute association edges represent the similarity of product attributes; Analyze the behavior sequence information in the customer behavior trajectory data and construct a behavior sequence edge between the customer behavior node and the customer node; wherein the behavior sequence edge represents the frequency or conversion rate of the behavior; Analyzing the product combination sales data between the customer transaction data and the product attribute data, and constructing a combination association edge between the product combination node and the product node; wherein the combination association edge represents the contribution of the product to the combination sales; Analyze the environmental impact information between the environmental data, the customer transaction data, and the product attribute data, and construct environmental impact edges between the environmental node and the customer node and the product node; wherein the environmental impact edges represent the degree of influence of environmental factors on customer behavior or product sales; Analyze seasonal variation information among the environmental data, the customer transaction data, and the product attribute data, and construct an environmental adjustment edge between the environmental node and the product combination node; wherein the environmental adjustment edge represents the impact of seasonal factors on product combination sales.

5. The method according to claim 2, characterized in that The operations of establishing a cross-layer interaction mechanism for the preliminary heterogeneous graph network and performing multi-granularity attribute fusion processing to generate the final heterogeneous graph network include: Establishing a cross-layer interaction mechanism between the basic physical layer, the behavioral semantic layer, and the environmental association layer in the preliminary heterogeneous graph network through hyperedges or meta-paths; fusing the static attributes of the customer node and the dynamic attributes of the customer behavior node in the preliminary heterogeneous graph network to generate comprehensive customer feature attributes, and adding the comprehensive customer feature attributes to the node attributes of the customer node; Determining semantic association feature attributes between products based on the node attributes of the product node and the node attributes of the product combination node, and adding the semantic association feature attributes to the node attributes of the product node; and According to the node attributes of the environment nodes, the edge weights in the preliminary heterogeneous graph network are dynamically adjusted to obtain a final heterogeneous graph network.

6. A storage medium, characterized in that The storage medium includes a stored program, wherein when the program is run, the processor executes the method according to any one of claims 1 to 5.

7. A data annotation device based on graph structure and community discovery, characterized in that: include: A data acquisition module is used to acquire multi-dimensional data of the customer group to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; a graph construction module for constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer, and an environmental association layer; the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products; the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products; and the environmental association layer is used to characterize the dynamic impact of external environmental factors on the interactive behavior between products and customers; and A labeling module is configured to divide the customer group into multiple customer communities based on the heterogeneous graph network through a community discovery algorithm, and generate a hierarchical label tree for each customer community; wherein the hierarchical label tree includes a primary label, a secondary label, and a tertiary label, wherein the primary label is constructed based on the interaction pattern between the basic physical layer and the environmental association layer and serves as the top node of the hierarchical label tree; the secondary label is constructed based on the potential behavior pattern of customers in the customer community in the behavior semantic layer and serves as the middle node of the hierarchical label tree; the tertiary label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network and serves as the bottom node of the hierarchical label tree; The operation of dividing the customer group into a plurality of customer communities by using a community discovery algorithm based on the heterogeneous graph network includes: Performing community division based on transaction associations at the basic physical layer of the heterogeneous graph network to form initial communities centered around physical attributes of goods and transaction frequencies; Performing community optimization based on behavioral patterns in the behavioral semantic layer of the heterogeneous graph network, and adjusting the community boundaries of the initial community in combination with customer behavior sequences and product combination association characteristics; and Introducing dynamic environmental factors into the environmental association layer of the heterogeneous graph network, performing time-series sensitivity correction on the adjusted initial community, and obtaining the corresponding customer community; The operation of generating a hierarchical tag tree for each customer community includes: Determine the basic value attributes of each customer community based on the basic physical layer; determine the value fluctuation coefficient of each customer community based on the environmental association layer; and determine a primary tag for each customer community based on the basic value attributes and the value fluctuation coefficient as the top-level node of the hierarchical tag tree. Based on the behavioral semantic layer, determining the behavioral sequence information of customers in each customer community; based on the product nodes and product combination nodes in the heterogeneous graph network, determining the product purchase status of customers in each customer community; based on the behavioral sequence information and the product purchase status, determining a secondary label for each customer community as an intermediate layer node of the hierarchical label tree; and Based on the basic physical layer, the behavioral semantic layer, and the environmental association layer, the temporal characteristic indicators of customers in each customer community are calculated, and the third-level tags of each customer community are determined according to the temporal characteristic indicators as the bottom-level nodes of the hierarchical tag tree; wherein the temporal characteristic indicators include purchase cycle volatility and activity participation frequency change rate.

8. A data annotation device based on graph structure and community discovery, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Obtain multi-dimensional data of the customer group to be labeled; wherein the multi-dimensional data includes customer transaction data, product attribute data, customer behavior trajectory data and environmental data; Constructing a heterogeneous graph network based on the multi-dimensional data; wherein the heterogeneous graph network includes a basic physical layer, a behavioral semantic layer, and an environmental association layer, wherein the basic physical layer is used to characterize the transaction relationship between customers and products and the physical attributes of the products; the behavioral semantic layer is used to characterize the potential behavior patterns of customers and the semantic association characteristics between products; and the environmental association layer is used to characterize the dynamic impact of external environmental factors on the interactive behavior between products and customers; and Based on the heterogeneous graph network, the customer group is divided into multiple customer communities through a community discovery algorithm, and a hierarchical label tree is generated for each customer community; wherein the hierarchical label tree includes a primary label, a secondary label, and a tertiary label, wherein the primary label is constructed based on the interaction pattern of the basic physical layer and the environmental association layer, and serves as the top node of the hierarchical label tree; the secondary label is constructed based on the potential behavior pattern of customers in the customer community in the behavior semantic layer, and serves as the middle node of the hierarchical label tree; the tertiary label is constructed based on the dynamic trajectory time series characteristics of customers in the customer community in the heterogeneous graph network, and serves as the bottom node of the hierarchical label tree; The operation of dividing the customer group into a plurality of customer communities by using a community discovery algorithm based on the heterogeneous graph network includes: Performing community division based on transaction associations at the basic physical layer of the heterogeneous graph network to form initial communities centered around physical attributes of goods and transaction frequencies; Performing community optimization based on behavioral patterns in the behavioral semantic layer of the heterogeneous graph network, and adjusting the community boundaries of the initial community in combination with customer behavior sequences and product combination association characteristics; and Introducing dynamic environmental factors into the environmental association layer of the heterogeneous graph network, performing time-series sensitivity correction on the adjusted initial community, and obtaining the corresponding customer community; The operation of generating a hierarchical tag tree for each customer community includes: Determine the basic value attributes of each customer community based on the basic physical layer; determine the value fluctuation coefficient of each customer community based on the environmental association layer; and determine a primary tag for each customer community based on the basic value attributes and the value fluctuation coefficient as the top-level node of the hierarchical tag tree. Based on the behavioral semantic layer, determining the behavioral sequence information of customers in each customer community; based on the product nodes and product combination nodes in the heterogeneous graph network, determining the product purchase status of customers in each customer community; based on the behavioral sequence information and the product purchase status, determining a secondary label for each customer community as an intermediate layer node of the hierarchical label tree; and Based on the basic physical layer, the behavioral semantic layer, and the environmental association layer, the temporal characteristic indicators of customers in each customer community are calculated, and the third-level tags of each customer community are determined according to the temporal characteristic indicators as the bottom-level nodes of the hierarchical tag tree; wherein the temporal characteristic indicators include purchase cycle volatility and activity participation frequency change rate.

Images