A Method and System for Constructing Industry Maps Based on Contextual Marketing

Abstract

This invention provides a method and system for constructing an industry graph based on scenario-based marketing, belonging to the field of industry graph construction technology. The method includes: processing media text to generate a text dataset using a BERT pre-trained model; processing scenario data to generate a behavioral dataset using a spatiotemporal calibration algorithm; aggregating these to generate an unstructured marketing dataset; combining the structured marketing dataset to generate a multimodal marketing scenario dataset; using a stream-batch integrated framework and entity disambiguation technology to process the multimodal dataset and construct a four-dimensional data lake; constructing a three-dimensional feature library and training a marketing scenario weight model; mapping the data lake to a graph database to form an industry chain topology; adjusting the strength of node relationships based on the weight model; learning the implicit features of nodes using the GraphSAGE algorithm; and generating a scenario-sensitive industry graph topology, thereby reflecting the relationships between industry entities in different marketing scenarios in real time and improving the accuracy of the industry graph in marketing decisions.

Description

Technical Field

[0001] This invention relates to the field of industry map construction technology, and in particular to an industry map construction method and system based on scenario-based marketing. Background Technology

[0002] Traditional industry mapping methods heavily rely on structured data such as enterprise registration information and industry reports, resulting in significant technical bottlenecks: they only include limited dimensions such as basic enterprise information, data updates are lagging and the dimensions are singular, failing to accurately reflect real-time supply and demand changes in the industry chain, and the coordination between human intervention and dynamic updates is poor; they lack a complete scenario-sensitive mechanism and lack deep integration with unstructured data such as social media sentiment and user behavior patterns, causing existing industry maps to fail to accurately reflect market dynamics, leading to a disconnect between marketing decisions and actual market demand; and they use fixed-weight algorithms to integrate data, making it difficult to adapt to the differentiated needs of different marketing scenarios.

[0003] Therefore, it is necessary to provide a method and system for constructing an industry map based on scenario-based marketing to solve the above-mentioned technical problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for constructing an industry map based on scenario-based marketing. This system solves the problems in existing technologies where data updates are lagging and have a single dimension, failing to accurately reflect real-time supply and demand changes in the industry chain, and lacking the unstructured data fusion capabilities required for scenario-based marketing, leading to a disconnect between marketing decisions and market demand.

[0005] This invention provides a method for constructing an industry map based on scenario-based marketing, the method comprising:

[0006] Social media text is obtained through distributed crawling, and after being processed by a BERT pre-trained model, a marketing text dataset with semantic labels is generated. At the same time, offline scene data is collected through IoT devices, and after being processed by a spatiotemporal calibration algorithm, a marketing behavior dataset with spatiotemporal labels is generated. The marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels are combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated.

[0007] The multimodal marketing scenario dataset is preprocessed using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology, and a four-dimensional dynamic data lake is constructed.

[0008] Acquire historical marketing scenario datasets and construct a 3D marketing scenario feature library. Train the 3D marketing scenario feature library with an LSTM neural network to obtain a marketing scenario weight model.

[0009] The four-dimensional dynamic data lake is mapped to the Neo4j graph database to form an industry chain topology. The strength of the node relationships in the industry chain topology is adjusted based on the marketing scenario weight model. A graph neural network is introduced and the implicit features of the nodes in the industry chain topology are learned through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology.

[0010] Preferably, the process involves acquiring social media text via distributed crawlers, processing it using a BERT pre-trained model to generate a marketing text dataset with semantic tags, and simultaneously collecting offline scenario data via IoT devices, processing it using a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal tags. The semantically tagged marketing text dataset and the spatiotemporally tagged marketing behavior dataset are then combined to obtain an unstructured marketing dataset. This unstructured marketing dataset is then combined with the structured marketing dataset to generate a multimodal marketing scenario dataset, specifically including:

[0011] The social media text is acquired in real time through the distributed crawler cluster, and the BERT pre-trained model is used to perform semantic encoding on the social media text. The industry chain entities in the social media text are extracted through named entity recognition technology, and the relationship between the industry chain entities is constructed by dependency parsing. The industry chain entities and their relationships are summarized to generate the marketing text dataset with semantic tags.

[0012] Simultaneously, the IoT device is connected to the edge computing gateway to collect the offline scene data. The offline scene data is then geofence matched and timestamp calibrated using the spatiotemporal calibration algorithm. This involves mapping the offline scene data to spatial entity coordinates and labeling the collection time granularity of the offline scene data, thus forming the marketing behavior dataset with spatiotemporal tags.

[0013] The unstructured marketing dataset is obtained by summing the marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels, and the structured marketing dataset is obtained by synchronously accessing the ERP system.

[0014] The unstructured marketing dataset and the structured marketing dataset are combined to form the multimodal marketing scenario dataset.

[0015] Preferably, the step of preprocessing the multimodal marketing scenario dataset and constructing a four-dimensional dynamic data lake using the Apache Flink stream-batch processing framework and knowledge graph entity disambiguation technology specifically includes:

[0016] The Apache Flink streaming and batch processing framework is used to perform real-time event processing on the unstructured marketing dataset, while micro-batch processing is performed on the structured marketing dataset.

[0017] The knowledge graph entity disambiguation technology is used to construct an industry chain entity dictionary based on the knowledge graph. A deep semantic matching algorithm is then used to disambiguate the unstructured marketing dataset and the structured marketing data based on the industry chain entity dictionary and to establish a unified entity identifier.

[0018] The multimodal marketing scenario dataset is mapped to the four-dimensional dynamic data lake based on a distributed storage architecture.

[0019] Preferably, the step of acquiring historical marketing scenario datasets and constructing a 3D marketing scenario feature library, and then training the 3D marketing scenario feature library to obtain a marketing scenario weight model using an LSTM neural network, specifically includes:

[0020] Obtain the historical marketing scenario dataset, and extract the time marketing features, spatial marketing features, and behavioral marketing features from the historical marketing scenario dataset to construct the three-dimensional marketing scenario feature library;

[0021] The LSTM neural network is used to train the 3D marketing scene feature library, and the marketing weight vectors corresponding to the marketing scene features in the 3D marketing scene feature library are matched to generate the marketing scene weight model.

[0022] Preferably, the step of mapping the four-dimensional dynamic data lake to the Neo4j graph database to form an industry chain topology, adjusting the node relationship strength in the industry chain topology based on the marketing scenario weight model, introducing a graph neural network and learning the implicit features of the nodes in the industry chain topology through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology specifically includes:

[0023] The four-dimensional dynamic data lake is mapped to the Neo4j graph database to construct the industry chain topology. The enterprise ID, product ID, and user ID in the four-dimensional dynamic data lake are mapped to nodes in the industry chain topology. The supply and demand relationship and public opinion relationship in the four-dimensional dynamic data lake are mapped to edges in the industry chain topology. The spatiotemporal attributes in the four-dimensional dynamic data lake are used as labels for nodes or edges in the industry chain topology.

[0024] The edge weights, i.e. the node relationship strength, are adjusted in real time based on the marketing scenario weight model.

[0025] The graph neural network is introduced, and the GraphSAGE algorithm is used to learn the embedded representation of the nodes. The features of adjacent nodes are weighted and aggregated in combination with the node relationship strength to obtain the hidden features of the nodes.

[0026] The industry chain topology is recalculated based on the node relationship strength and the node implicit features, and the key node clusters in the updated industry chain topology are identified by the community discovery algorithm to generate the scenario-sensitive industry graph topology.

[0027] Preferably, the real-time adjustment of edge weights based on the marketing scenario weight model specifically includes:

[0028] The FastText text classification model is used to extract keywords and perform semantic classification on the current marketing scenario description text in the four-dimensional dynamic data lake to determine the corresponding current marketing scenario type and extract the corresponding current marketing scenario features.

[0029] Based on the marketing scenario weight model, the current marketing weight vector corresponding to the current marketing scenario feature is matched, which is the edge weight.

[0030] Preferably, the entity set in the four-dimensional dynamic data lake is defined as follows: ,in Let i represent the i-th entity, i.e., enterprise ID, product ID, or user ID, and n represent the total number of entities in the four-dimensional dynamic data lake. The entity set E is mapped to the node set in the industry chain topology G. ,and And map() represents the mapping function from entity to node; the set of edges in the industry chain topology G is defined as ,in Represents the i-th node and the j-th node The relationship types are defined as follows: the spatiotemporal attribute set in the four-dimensional dynamic data lake is... ,in Indicates time attribute, Indicates spatial attributes;

[0031] The formula for calculating the hidden features of the nodes is as follows:

[0032] In the formula, Represents a node The hidden features of nodes in the l-th layer of a graph neural network; Indicates the activation function; This represents the feature aggregation function of the l-th layer; Represents a node Hidden features of nodes in the (l-1)th layer of a graph neural network; This represents the adjusted edge weight, i.e., the node weight. and The strength of the node relationships between them; Represents a node The set of neighboring nodes; Represents a node Hidden features of nodes in layer l-1 of a graph neural network.

[0033] Preferably, the step of recalculating the industry chain topology based on the node relationship strength and the node implicit characteristics specifically includes:

[0034] Based on the node relationship strength and the node implicit features, the similarity function is constructed as follows:

[0035] In the formula, Represents a node and nodes The similarity function; L represents the last layer of the graph neural network; Represents a node Hidden features of nodes in the last layer of a graph neural network; Represents a node Hidden features of nodes in the last layer of a graph neural network; Representing the hidden features of nodes The L2 norm; Representing the hidden features of nodes The L2 norm; Represents a node Hidden features of nodes With nodes Hidden features of nodes Cosine similarity; The empirical coefficient representing the amplification factor of the node relationship strength. This represents the amplification factor for the strength of the node relationship. ;

[0036] Set similarity threshold ,when At that time, retain the edge Otherwise delete the edge. The industry chain topology is recalculated to obtain the updated industry chain topology. ;

[0037] The step of identifying key node clusters in the updated industry chain topology using a community discovery algorithm to generate the scenario-sensitive industry graph topology specifically includes:

[0038] The updated industry chain topology Each node in the process forms a community. ,in Indicates the i-th community;

[0039] The community detection algorithm is used to calculate the communities respectively. and neighboring communities The community module level is and and calculate the community With neighboring communities The modularity of the merged new community is The community modularity gain is obtained as ;

[0040] If the community modularity gain Then merge the communities. and neighboring communities Otherwise, maintain the community. and neighboring communities independent;

[0041] The final set of communities is obtained by iterating through all adjacent community pairs. , where p represents the final community set The total number of communities in the area;

[0042] Calculate the final community set Each community Community Importance Score ,in Importance score for all communities Arrange the communities from largest to smallest, select the top M communities corresponding to their importance scores to form a key community cluster, i.e., the key node cluster, and generate the scenario-sensitive industry map topology structure.

[0043] A method for constructing an industry map based on scenario-based marketing also includes setting up a WebGL-based 3D visualization engine to support users in performing spatiotemporal slicing of the scenario-sensitive industry map topology using geofences and timestamps.

[0044] An improved PageRank algorithm is used to calculate the importance of nodes in the scenario-sensitive industry graph topology, and the DBSCAN clustering algorithm is used to identify key marketing touchpoints based on the importance of the nodes and generate marketing decision suggestions. The key marketing touchpoints are marked in the form of a heat map.

[0045] An interactive analytics interface is set up to allow users to click on the key marketing touchpoints and view the marketing decision recommendations.

[0046] An industry map construction system based on scenario-based marketing, the system comprising:

[0047] The data acquisition module is used to acquire social media text through distributed crawlers, process it with a BERT pre-trained model to generate a marketing text dataset with semantic labels, and collect offline scene data through IoT devices, process it with a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal labels. The marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels are combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated.

[0048] The data processing module is used to preprocess the multimodal marketing scenario dataset and construct a four-dimensional dynamic data lake using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology.

[0049] The model building module is used to acquire historical marketing scenario datasets and build a three-dimensional marketing scenario feature library. The marketing scenario weight model is obtained by training the three-dimensional marketing scenario feature library through an LSTM neural network.

[0050] The graph construction module is used to map the four-dimensional dynamic data lake to the Neo4j graph database to form an industry chain topology. Based on the marketing scenario weight model, the strength of the node relationships in the industry chain topology is adjusted. A graph neural network is introduced and the implicit features of the nodes in the industry chain topology are learned through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology.

[0051] Compared with related technologies, the industry map construction method and system based on scenario-based marketing provided by this invention has the following beneficial effects:

[0052] This invention acquires social media text through distributed web crawlers, processes it using a BERT pre-trained model to generate a marketing text dataset with semantic tags, and simultaneously collects offline scene data through IoT devices, processes it using a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal tags. The semantically tagged marketing text dataset and the spatiotemporally tagged marketing behavior dataset are then combined to obtain an unstructured marketing dataset. Finally, the structured marketing dataset is combined to generate a multimodal marketing scenario dataset. The invention utilizes Apache... The Flink streaming and batch processing framework and knowledge graph entity disambiguation technology are used to preprocess multimodal marketing scenario datasets and construct a four-dimensional dynamic data lake. Historical marketing scenario datasets are acquired and a three-dimensional marketing scenario feature library is built. An LSTM neural network is used to train the three-dimensional marketing scenario feature library to obtain a marketing scenario weight model. The four-dimensional dynamic data lake is mapped to the Neo4j graph database to form an industry chain topology. Based on the marketing scenario weight model, the strength of node relationships in the industry chain topology is adjusted. A graph neural network is introduced, and the GraphSAGE algorithm is used to learn the implicit features of nodes in the industry chain topology, generating a scenario-sensitive industry graph topology. This allows for the construction of scenario-sensitive industry graphs to reflect the relationships between industry entities in different marketing scenarios in real time, thereby improving the effectiveness and accuracy of the industry graph in marketing decisions.

[0053] This invention integrates semantic data from social media texts and spatiotemporal behavioral data collected from IoT devices, combined with structured data from ERP systems, to generate a multimodal marketing scenario dataset, enabling the construction of industry graphs based on a more comprehensive and richer data foundation. The method utilizes the Apache Flink streaming-batch processing framework to achieve real-time processing of unstructured data and micro-batch processing of structured data, ensuring the timeliness and efficiency of data processing. By constructing a 3D marketing scenario feature library and training a marketing scenario weight model using an LSTM neural network, this invention can adjust the strength of node relationships in the industry chain topology in real time according to different marketing scenarios, making the industry graph scenario-sensitive and more accurately reflecting the relationships between industry entities under different marketing scenarios. This invention combines graph neural networks and the GraphSAGE algorithm to learn the implicit features of nodes, and optimizes the industry chain topology through similarity functions and community detection algorithms, generating a more accurate scenario-sensitive industry graph. Furthermore, this invention provides users with intuitive and effective marketing decision support through a 3D visualization engine, node importance calculation, key marketing touchpoint identification, and interactive analysis interfaces, improving the scientific nature and accuracy of marketing decisions. Attached Figure Description

[0054] Figure 1 A flowchart illustrating an industry map construction method based on scenario-based marketing, provided as an embodiment of the present invention;

[0055] Figure 2 A system block diagram of an industry map construction system based on scenario-based marketing provided for embodiments of the present invention;

[0056] Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] like Figure 1 The diagram shown is a flowchart of an industry map construction method based on scenario-based marketing provided by an embodiment of the present invention. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. Steps S1 to S4 are detailed as follows:

[0059] S1. Social media text is obtained through distributed crawling, and after being processed by a BERT pre-trained model, a marketing text dataset with semantic labels is generated. At the same time, offline scene data is collected through IoT devices, and after being processed by a spatiotemporal calibration algorithm, a marketing behavior dataset with spatiotemporal labels is generated. The marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels are combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated.

[0060] Distributed web crawling is a technique that uses multiple computer nodes to crawl network data in parallel. It leverages a distributed architecture to improve data collection efficiency and is used to acquire large-scale social media text, such as comments, posts, and messages, in real time, solving the problems of slow crawling speed and susceptibility to blocking by single nodes. The BERT pre-trained model refers to a pre-trained language model based on the Transformer architecture. Trained through a bidirectional masked language model, it can capture deep semantics of text. In the method of this invention, it is used to perform named entity recognition and relation extraction on social media text, generating a marketing text dataset with semantic labels. A marketing text dataset with semantic labels refers to a dataset containing explicit semantic annotations generated after semantic parsing of social media text.

[0061] IoT (Internet of Things) devices refer to physical devices with network connectivity, such as Wi-Fi probes, thermal sensors, and RFID tags. They are used to collect offline scene data, such as foot traffic in shopping malls and user movement trajectories. Spatiotemporal calibration algorithms refer to algorithms that perform timestamp calibration and geospatial matching on data collected by IoT devices, such as R-tree spatiotemporal indexes. These algorithms can map raw sensor data to specific geofences and time granularities, generating marketing behavior datasets with spatiotemporal labels. A geofence might be the coordinates of a shopping mall, and the time granularity might be 10 minutes per instance. A marketing behavior dataset with spatiotemporal labels refers to a dataset generated by spatiotemporally calibrating offline scene data collected through IoT devices, and including time and spatial dimension annotations.

[0062] Unstructured marketing datasets refer to data collections that do not follow a fixed data model, including social media texts and user behavior trajectories. These datasets have flexible formats. In this invention, semantic parsing and spatiotemporal calibration are used to transform unstructured data into processable, tagged data. Structured marketing datasets refer to data with a defined data model and format, such as order data from ERP (Enterprise Resource Planning) systems and structured statistical data from industry reports. These, combined with unstructured data, form multimodal marketing scenario datasets to cover all dimensions of information needed for marketing decisions. Multimodal marketing scenario datasets are comprehensive datasets used to integrate multiple data modalities, covering all dimensions of information within a marketing scenario.

[0063] In practical applications, distributed web crawling technology can be used to batch acquire user-generated text content from social media platforms. Using a BERT pre-trained model, semantic parsing of the text is performed, and natural language processing techniques such as named entity recognition and relation extraction are employed to generate a structured marketing text dataset with entity tags, relation annotations, and sentiment analysis. Simultaneously, user behavior data from offline scenarios can be collected using IoT devices. Spatiotemporal calibration algorithms are used to synchronize timestamps and perform geofencing matching on this data, forming a marketing behavior dataset containing location coordinates and time dimensions. Finally, these two types of unstructured data are integrated with structured data such as order data and industry analysis reports from enterprise resource planning systems to create a multimodal marketing scenario dataset covering semantic information, spatiotemporal behavior, and business statistics.

[0064] S2, using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology to preprocess the multimodal marketing scenario dataset and construct a four-dimensional dynamic data lake;

[0065] It's understandable that the Apache Flink streaming and batch processing framework refers to an open-source big data processing framework that supports unified processing of real-time streaming data and batch data. Real-time streaming data includes things like live stream comments, while batch data includes things like daily reports. Knowledge graph entity disambiguation technology can solve the entity aliasing problem in multi-source data through deep semantic matching algorithms and builds an entity dictionary based on the knowledge graph, assigning unified identifiers to entities from different data sources to achieve entity alignment across modal data and ensure the accuracy of data fusion. A four-dimensional dynamic data lake refers to a distributed data storage architecture built according to four dimensions: entity, relationship, spatiotemporal, and attribute, supporting real-time incremental data updates and data lineage tracing.

[0066] It should be noted that the Apache Flink streaming and batch processing framework can be used to uniformly clean and standardize real-time live stream comments, device sensor data, and batch-imported daily sales reports. Knowledge graph entity disambiguation technology is used to solve the semantic consistency problem of company names and product aliases across different data sources, thereby constructing a four-dimensional dynamic data lake. This data lake continuously integrates new data through an incremental learning mechanism, ensuring that data update latency is kept within a short cycle.

[0067] S3, acquire historical marketing scenario dataset and construct a three-dimensional marketing scenario feature library, and train the three-dimensional marketing scenario feature library through LSTM neural network to obtain a marketing scenario weight model;

[0068] It should be noted that the historical marketing scenario dataset refers to a comprehensive collection of data generated from past marketing activities, encompassing business data, user behavior data, and marketing effectiveness data across different marketing scenarios. The three-dimensional marketing scenario feature library is a feature set built upon historical marketing data, including time, space, and behavioral dimensions. It reflects the core attributes of different marketing scenarios through three-dimensional feature vectors. These include time dimensions such as promotional cycle stages, spatial dimensions such as regional consumer spending power indices, and behavioral dimensions such as user conversion rates. Marketing scenarios include new product launches, holiday promotions, and regional expansion. The LSTM neural network, or Long Short-Term Memory network, is used in this invention to train the three-dimensional feature library, learning the relationship between scenario features and weight mappings to generate a scenario-sensitive marketing scenario weight model. The marketing scenario weight model is an algorithm model trained on the historical marketing scenario dataset, used to dynamically adjust the calculated weights of each data dimension in the industry map according to the current marketing scenario type.

[0069] In practical applications, full-scale data from different scenarios across historical marketing campaigns can be collected to construct a three-dimensional feature library encompassing time, space, and behavioral dimensions. Then, an LSTM can be used to train the feature library, learning the mapping relationship between features and data weights for different scenarios, generating a marketing scenario weight model. This model can automatically adjust the calculated weights of dimensions such as supply and demand relationships and competitive landscape in supply chain analysis based on real-time scenario types, achieving dynamic adaptation of data weights.

[0070] S4. The four-dimensional dynamic data lake is mapped to the Neo4j graph database to form an industry chain topology. The strength of the node relationships in the industry chain topology is adjusted based on the marketing scenario weight model. A graph neural network is introduced and the implicit features of the nodes in the industry chain topology are learned through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology.

[0071] Neo4j graph database refers to a NoSQL database based on a graph model. It uses nodes, edges, and attributes to represent data and relationships, making it suitable for storing and querying complex relationships such as those in the industry chain. It can map a four-dimensional data lake into a graph structure, supporting efficient graph topology analysis. Node relationship strength refers to the weight of edges in the industry chain topology, reflecting the tightness of the connections between nodes. Graph Neural Networks (GNNs) are neural networks that operate directly on graph structures. The GraphSAGE algorithm is a classic algorithm in graph neural networks, which solves the training efficiency problem of large-scale graph data by sampling neighboring nodes and aggregating their features. Node latent features refer to abstract feature vectors containing deep semantic information generated after multi-layer aggregation learning of node features in the industry chain topology through graph neural networks. Scenario-sensitive industry graph topology refers to an industry chain graph used to dynamically adjust the node relationship strength and topological connections according to the current marketing scenario. It can adaptively present key nodes and core connections highly relevant to the marketing scenario.

[0072] Furthermore, entities, relationships, and spatiotemporal attributes in the four-dimensional dynamic data lake can be mapped to the Neo4j graph database to construct an industry chain topology composed of nodes and edges. Then, edge weights in the graph can be calibrated in real time based on a marketing scenario weight model, for example, enhancing the interaction edge weights between stores and users in promotional scenarios. Graph neural networks can be introduced, employing the GraphSAGE algorithm for multi-layer learning of node features. By weighted aggregation of scenario-sensitive features of neighboring nodes, latent node representations containing deep semantics can be generated, such as store features that integrate regional customer flow and user reviews. Finally, by combining latent node features and dynamic edge weights, a community detection algorithm can be used to recalculate the industry chain topology, identifying highly correlated key node clusters, such as core promotional business districts, generating an industry graph topology directly serving the current marketing scenario, thus upgrading from data display to decision support.

[0073] In the specific implementation process, social media text is obtained through distributed web crawling, processed by a BERT pre-trained model to generate a marketing text dataset with semantic labels, and simultaneously, offline scene data is collected through IoT devices, processed by a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal labels. The semantically labeled marketing text dataset and the spatiotemporally labeled marketing behavior dataset are then combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated, specifically including:

[0074] The social media text is acquired in real time through the distributed crawler cluster, and the BERT pre-trained model is used to perform semantic encoding on the social media text. The industry chain entities in the social media text are extracted through named entity recognition technology, and the relationship between the industry chain entities is constructed by dependency parsing. The industry chain entities and their relationships are summarized to generate the marketing text dataset with semantic tags.

[0075] Simultaneously, the IoT device is connected to the edge computing gateway to collect the offline scene data. The offline scene data is then geofence matched and timestamp calibrated using the spatiotemporal calibration algorithm. This involves mapping the offline scene data to spatial entity coordinates and labeling the collection time granularity of the offline scene data, thus forming the marketing behavior dataset with spatiotemporal tags.

[0076] The unstructured marketing dataset is obtained by summing the marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels, and the structured marketing dataset is obtained by synchronously accessing the ERP system.

[0077] The unstructured marketing dataset and the structured marketing dataset are combined to form the multimodal marketing scenario dataset.

[0078] This approach employs a distributed crawler cluster architecture, utilizing a multi-node parallel computing strategy to capture user-generated content from social media platforms in real time, forming a large-scale text data collection network. By leveraging a BERT pre-trained model for deep semantic encoding of the original text, and utilizing named entity recognition technology to accurately extract industry chain entities from the text, covering semantic units such as enterprise entities, product types, and market behaviors, the approach further utilizes dependency parsing technology to uncover grammatical dependencies and semantic logical connections between entities. This constructs structured association networks such as supply relationships, competitive relationships, and consumer behavior, transforming unstructured text into a semantically labeled dataset containing entity labels and relationship links, providing textual semantic support for industry mapping.

[0079] Furthermore, an IoT device access hub can be built using edge computing gateways to achieve unified management and data collection of offline sensing devices, including wireless probes, thermal sensors, and RFID tags. Then, a spatiotemporal calibration algorithm can be used to perform dual processing on the collected raw sensor data: spatially, geofencing matching technology maps sensor data to specific physical coordinate entities, such as shopping mall areas or store locations; temporally, the data is timestamped to standardize the data and unify the collection time granularity, such as sampling at fixed time intervals. Ultimately, a marketing behavior dataset with precise spatial location and time stamps can be generated, enabling data-driven representation of user behavior and providing spatiotemporal dynamic information of offline scenarios for the industry map.

[0080] Furthermore, the semantically parsed marketing text dataset can be heterogeneously integrated with the spatiotemporally calibrated marketing behavior dataset to form an unstructured marketing dataset. This integration process involves data cleaning, format standardization, and metadata annotation, resolving the format consistency issue of unstructured data from different sources. Specifically, semantic entities in the text data are associated with physical entities in the spatiotemporal data through entity alignment techniques. For example, matching store names mentioned in social media with customer flow data of those stores collected by IoT devices creates cross-modal entity association chains, enhancing the integration value of unstructured data.

[0081] Finally, business data from the ERP (Enterprise Resource Planning) system, including typical structured data such as order records, inventory information, and sales reports, can be synchronously integrated with the unstructured marketing dataset for deep fusion. During the fusion process, multi-source data mapping technology is employed to establish correspondences between business entities in the structured data and semantic and spatiotemporal entities in the unstructured data, constructing a four-dimensional data model encompassing entities, relationships, spatiotemporal dimensions, and attributes. The multimodal marketing scenario dataset formed through this model integrates multi-dimensional information including textual semantics, physical spatiotemporal dimensions, and business statistics, breaking through the dimensional limitations of traditional single-structured data and providing a comprehensive data foundation for scenario-based industry mapping.

[0082] The process of preprocessing the multimodal marketing scenario dataset and constructing a four-dimensional dynamic data lake using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology specifically includes:

[0083] The Apache Flink streaming and batch processing framework is used to perform real-time event processing on the unstructured marketing dataset, while micro-batch processing is performed on the structured marketing dataset.

[0084] The knowledge graph entity disambiguation technology is used to construct an industry chain entity dictionary based on the knowledge graph. A deep semantic matching algorithm is then used to disambiguate the unstructured marketing dataset and the structured marketing data based on the industry chain entity dictionary and to establish a unified entity identifier.

[0085] The multimodal marketing scenario dataset is mapped to the four-dimensional dynamic data lake based on a distributed storage architecture.

[0086] Based on the Apache Flink streaming and batch processing framework, unstructured marketing data can be processed using a real-time event-driven model. For real-time data streams such as social media sentiment streams and IoT device sensor data, continuous processing is performed based on event time semantics to ensure data timeliness. For structured marketing data, micro-batch processing can be implemented, dividing the data into sub-batches for computation at fixed time intervals, thus balancing processing efficiency and resource consumption. This architecture achieves a unified processing paradigm for different modalities of data and solves the challenge of collaborative computation between real-time and batch data.

[0087] Then, a supply chain entity dictionary can be constructed based on the knowledge graph, integrating entity representations from multiple data sources to form a standardized entity knowledge base. Furthermore, deep semantic matching algorithms can be used to semantically align semantic entities in unstructured data with business entities in structured data. Semantic entities include company aliases mentioned in social media texts, while business entities include registered company names in ERP systems. In addition, a globally unique unified entity identifier can be assigned to each entity, establishing entity association links across modal data to ensure consistency and traceability of data from different sources.

[0088] Furthermore, based on a distributed storage architecture, the preprocessed multimodal marketing scenario dataset can be mapped to a dynamic data lake encompassing four dimensions: entity, relationship, time, and space. The entity dimension integrates various entities within the industry chain, the relationship dimension depicts business connections between entities, the time dimension records the data generation sequence, and the spatial dimension labels physical location attributes. Through this four-dimensional dynamic data lake, centralized data storage and management can be achieved, ensuring incremental updates and version control. This provides standardized and real-time data support for scenario-based industry mapping, significantly improving data integration efficiency and analysis quality.

[0089] The process of acquiring historical marketing scenario datasets and constructing a 3D marketing scenario feature library, and then training the 3D marketing scenario feature library using an LSTM neural network to obtain a marketing scenario weight model, specifically includes:

[0090] Obtain the historical marketing scenario dataset, and extract the time marketing features, spatial marketing features, and behavioral marketing features from the historical marketing scenario dataset to construct the three-dimensional marketing scenario feature library;

[0091] The LSTM neural network is used to train the 3D marketing scene feature library, and the marketing weight vectors corresponding to the marketing scene features in the 3D marketing scene feature library are matched to generate the marketing scene weight model.

[0092] This involves collecting full data from historical marketing campaigns, encompassing business data, user behavior data, and market feedback data across different marketing scenarios, forming a historical marketing scenario dataset. From this dataset, three core feature dimensions can be systematically extracted: In terms of time, this includes extracting temporal features such as the cycle stages of marketing campaigns and holiday attributes; in terms of space, it includes mining geographical attributes such as regional consumer spending power indices and the heat distribution of business districts; and in terms of behavior, it includes analyzing behavioral patterns such as user conversion rates, interaction frequency, and consumption preferences. These multi-dimensional features are then structurally integrated to construct a three-dimensional marketing scenario feature library containing time, space, and behavioral elements, providing standardized data samples for subsequent model training.

[0093] Furthermore, an LSTM neural network can be used as the core training algorithm. Its unique gating mechanism effectively handles long-sequence dependencies in historical marketing data, avoiding the gradient vanishing problem of traditional recurrent neural networks. A 3D marketing scenario feature library is input into the LSTM neural network as training samples. The network parameters are optimized through backpropagation to learn the mapping relationship between different marketing scenario features and data weights. During model training, a corresponding marketing weight vector is matched for each combination of scenario features. This vector contains weight coefficients for dimensions such as supply and demand, competitive landscape, and public opinion intensity, used to quantify the importance of each data dimension in a specific marketing scenario.

[0094] After training with an LSTM neural network, a decision-making model can be formed that automatically identifies marketing scenario features and outputs corresponding weight vectors—the marketing scenario weight model. This model possesses scenario adaptability, quickly activating appropriate weight strategies based on the real-time marketing scenario type. It dynamically adjusts the calculated weights of each data dimension in industry graph analysis; for example, increasing the weight of geofencing data in regional promotion scenarios and strengthening the weight of public opinion data in new product promotion scenarios, thereby achieving precise adaptation of marketing decisions to market dynamics. This model overcomes the limitations of traditional fixed-weight algorithms, significantly improving the scenario sensitivity and decision-making effectiveness of industry graph analysis.

[0095] The process involves mapping the four-dimensional dynamic data lake to the Neo4j graph database to form an industry chain topology. Based on the marketing scenario weight model, the strength of node relationships in the industry chain topology is adjusted. A graph neural network is introduced, and the GraphSAGE algorithm is used to learn the implicit features of nodes in the industry chain topology, generating a scenario-sensitive industry graph topology. Specifically, this includes:

[0096] The four-dimensional dynamic data lake is mapped to the Neo4j graph database to construct the industry chain topology. The enterprise ID, product ID, and user ID in the four-dimensional dynamic data lake are mapped to nodes in the industry chain topology. The supply and demand relationship and public opinion relationship in the four-dimensional dynamic data lake are mapped to edges in the industry chain topology. The spatiotemporal attributes in the four-dimensional dynamic data lake are used as labels for nodes or edges in the industry chain topology.

[0097] The edge weights, i.e. the node relationship strength, are adjusted in real time based on the marketing scenario weight model.

[0098] The graph neural network is introduced, and the GraphSAGE algorithm is used to learn the embedded representation of the nodes. The features of adjacent nodes are weighted and aggregated in combination with the node relationship strength to obtain the hidden features of the nodes.

[0099] The industry chain topology is recalculated based on the node relationship strength and the node implicit features, and the key node clusters in the updated industry chain topology are identified by the community discovery algorithm to generate the scenario-sensitive industry graph topology.

[0100] In practical applications, multimodal data from a four-dimensional dynamic data lake can be mapped to the Neo4j graph database to construct a visualized industry chain topology. Specifically, enterprise identifiers, product identifiers, and user identifiers from the four-dimensional dynamic data lake serve as the basic nodes of the industry chain topology, representing the subjects, objects, and actors in the industry chain, respectively. Business connections such as supply and demand relationships and public opinion correlations from the four-dimensional dynamic data lake are mapped as edges of the industry chain topology to characterize the business links and semantic relationships between entities. Simultaneously, the temporal and spatial attributes of the industry chain topology are used as labels for nodes or edges, embedding the spatiotemporal dimensions into the industry chain topology. This mapping process constructs a three-dimensional graph model containing entity nodes, relationship edges, and spatiotemporal attributes, providing a structured visualization carrier for industry chain analysis.

[0101] Then, based on the marketing scenario weight model, the edge weights in the industry chain topology can be calibrated in real time. This model is trained on historical marketing data and can automatically generate corresponding weight vectors according to the current marketing scenario type. During the weight adjustment process, the weight values ​​of specific types of edges are dynamically increased or decreased for different scenario characteristics, such as time period, geographical range, and user behavior. For example, in a holiday promotion scenario, the weight of the interaction edge between store nodes and user nodes is strengthened, while the weight of the cross-regional supply chain edge is weakened, so that the edge weight distribution of the industry chain topology can accurately reflect the business correlation strength under the marketing scenario requirements, achieving initial adaptation of the topology to the marketing scenario.

[0102] Furthermore, graph neural networks can be introduced, employing the GraphSAGE algorithm to learn embedded representations for nodes in the industry chain topology. This algorithm iterates through multiple layers of neural networks, weighting and aggregating the features of neighboring nodes based on the strength of node relationships, while simultaneously integrating the node's own initial features, such as company size and product reviews, to generate latent node features containing deep semantic information. Specifically, in each layer of the network, the algorithm weights the features of neighboring nodes according to edge weights, extracts common features through aggregation operations such as pooling and mean calculation, and then generates a higher-level abstract representation through nonlinear transformation. This process enables node features to capture complex relationships within the industry chain; for example, the latent features of a store node can be integrated with multi-dimensional information such as its geographical location, surrounding customer flow, and user reviews to form a scene-sensitive semantic representation of the node.

[0103] Finally, the industry chain topology can be recalculated and optimized based on the adjusted edge weights and the learned latent node features. Specifically, firstly, a similarity function is used to fuse latent node features and edge weights to measure the semantic association strength and scenario adaptability between nodes, thereby reconstructing node connections. Secondly, a community detection algorithm is used to modularize the new topology, identifying highly cohesive and loosely coupled node communities by maximizing the modularity index. Finally, based on the weighted degree and average similarity of nodes within a community, the importance score of each community is calculated, and communities with higher scores are selected as key node clusters, forming a scenario-sensitive industry graph topology. This topology can highlight the core business links and high-value touchpoints in the current marketing scenario, such as core promotional business districts and high-public-interest product clusters, providing visualized and precise guidance for marketing decisions.

[0104] The real-time adjustment of edge weights based on the marketing scenario weight model specifically includes:

[0105] The FastText text classification model is used to extract keywords and perform semantic classification on the current marketing scenario description text in the four-dimensional dynamic data lake to determine the corresponding current marketing scenario type and extract the corresponding current marketing scenario features.

[0106] Based on the marketing scenario weight model, the current marketing weight vector corresponding to the current marketing scenario feature is matched, which is the edge weight.

[0107] Specifically, the FastText text classification model can be used to perform semantic parsing of the current marketing scenario description text in the four-dimensional dynamic data lake. This model automatically identifies the current marketing scenario type by extracting keywords and semantically encoding the scenario description text, matching it with predefined scenario type labels. Based on the identified scenario type, further scene features in dimensions such as time, space, and behavior are extracted from the text to form a structured current marketing scenario feature vector.

[0108] Then, the extracted features of the current marketing scenario can be input into the marketing scenario weight model. Through matrix operations and nonlinear transformations, the corresponding current marketing weight vector is obtained. This weight vector contains weight coefficients for multiple data dimensions such as supply and demand, competitive landscape, and geographical distribution, used to quantify the importance of each dimension in the current scenario. For example, in a regional promotion scenario, the marketing scenario weight model will automatically increase the weight coefficient of the geographical distribution dimension and decrease the weight of irrelevant dimensions, so that the weight vector can accurately reflect the scenario requirements.

[0109] Finally, the matched marketing weight vector can be applied to adjust the edge weights of the industry chain graph. For each edge in the graph, the edge weight value is dynamically adjusted based on the corresponding coefficients in the weight vector and the data dimensions involved. Specifically, for edges highly relevant to the current scenario, their weight values ​​are strengthened according to the weight coefficients of the corresponding dimensions; for edges unrelated to the scenario, their weight values ​​are maintained or weakened. Through this mechanism, the edge weight distribution of the industry chain graph can adapt to changes in marketing scenarios in real time, making the graph topology more accurately reflect the strength of business connections in the current scenario, and providing dynamic data support for subsequent graph analysis and marketing decisions.

[0110] The entity set in the four-dimensional dynamic data lake is defined as follows: ,in Let i represent the i-th entity, i.e., enterprise ID, product ID, or user ID, and n represent the total number of entities in the four-dimensional dynamic data lake. The entity set E is mapped to the node set in the industry chain topology G. ,and And map() represents the mapping function from entity to node; the set of edges in the industry chain topology G is defined as ,in Represents the i-th node and the j-th node The relationship types are defined as follows: the spatiotemporal attribute set in the four-dimensional dynamic data lake is... ,in Indicates time attribute, Indicates spatial attributes;

[0111] The formula for calculating the hidden features of the nodes is as follows:

[0112] In the formula, Represents a node The hidden features of nodes in the l-th layer of a graph neural network; Indicates the activation function; This represents the feature aggregation function of the l-th layer; Represents a node Hidden features of nodes in the (l-1)th layer of a graph neural network; This represents the adjusted edge weight, i.e., the node weight. and The strength of the node relationships between them; Represents a node The set of neighboring nodes; Represents a node Hidden features of nodes in layer l-1 of a graph neural network.

[0113] The step of recalculating the industry chain topology based on the node relationship strength and the node implicit characteristics specifically includes:

[0114] Based on the node relationship strength and the node implicit features, the similarity function is constructed as follows:

[0115] In the formula, Represents a node and nodes The similarity function; L represents the last layer of the graph neural network; Represents a node Hidden features of nodes in the last layer of a graph neural network; Represents a node Hidden features of nodes in the last layer of a graph neural network; Representing the hidden features of nodes The L2 norm; Representing the hidden features of nodes The L2 norm; Represents a node Hidden features of nodes With nodes Hidden features of nodes Cosine similarity; The empirical coefficient representing the amplification factor of the node relationship strength. This represents the amplification factor for the strength of the node relationship. ;

[0116] Set similarity threshold ,when At that time, retain the edge Otherwise delete the edge. The industry chain topology is recalculated to obtain the updated industry chain topology. ;

[0117] The step of identifying key node clusters in the updated industry chain topology using a community discovery algorithm to generate the scenario-sensitive industry graph topology specifically includes:

[0118] The updated industry chain topology Each node in the process forms a community. ,in Indicates the i-th community;

[0119] The community detection algorithm is used to calculate the communities respectively. and neighboring communities The community module level is and and calculate the community With neighboring communities The modularity of the merged new community is The community modularity gain is obtained as ;

[0120] If the community modularity gain Then merge the communities. and neighboring communities Otherwise, maintain the community. and neighboring communities independent;

[0121] The final set of communities is obtained by iterating through all adjacent community pairs. , where p represents the final community set The total number of communities in the area;

[0122] Calculate the final community set Each community Community Importance Score ,in Importance score for all communities Arrange the communities from largest to smallest, select the top M communities corresponding to their importance scores to form a key community cluster, i.e., the key node cluster, and generate the scenario-sensitive industry map topology structure.

[0123] At the data modeling level, entities, relationships, and spatiotemporal attributes in a four-dimensional dynamic data lake can be mapped to nodes, edges, and labels in a graph database, constructing an industrial chain topology structure that includes entities such as enterprises, products, and users. This enables a unified graph model representation of multimodal data, solves the problem of traditional data fragmentation, and visualizes the industrial chain relationships.

[0124] At the feature learning level, the GraphSAGE algorithm of graph neural networks can be used to weight and aggregate node features, and the hidden features of nodes can be calculated by combining the edge weights after scene adjustment. This can capture the deep semantic relationship between nodes in the industry chain, which is more in line with the needs of marketing scenarios than traditional feature extraction methods and significantly improves feature representation capabilities.

[0125] At the topology optimization level, by reconstructing the industry chain topology through a similarity function that integrates node latent features and edge weights, irrelevant connections in the scenario can be automatically filtered out, while highly relevant edges are retained, allowing the graph topology to dynamically adapt to the current marketing scenario. Combined with community discovery algorithms to identify key node clusters, high-value marketing touchpoints can be accurately located and scenario-sensitive graphs can be generated. Real-world testing shows a significant improvement in touchpoint identification accuracy, effectively supporting precise resource allocation for scenario-based marketing.

[0126] A method for constructing an industry map based on scenario-based marketing also includes setting up a WebGL-based 3D visualization engine to support users in performing spatiotemporal slicing of the scenario-sensitive industry map topology using geofences and timestamps.

[0127] An improved PageRank algorithm is used to calculate the importance of nodes in the scenario-sensitive industry graph topology, and the DBSCAN clustering algorithm is used to identify key marketing touchpoints based on the importance of the nodes and generate marketing decision suggestions. The key marketing touchpoints are marked in the form of a heat map.

[0128] An interactive analytics interface is set up to allow users to click on the key marketing touchpoints and view the marketing decision recommendations.

[0129] First, a 3D visualization engine can be developed based on WebGL technology. This engine has powerful graphics rendering capabilities and allows users to define spatial ranges using geofencing tools and filter specific time periods using timestamp controls. This enables spatiotemporal slicing analysis of the industry map topology, allowing for dynamic filtering of industry chain data. Users can then focus on changes in business relationships within specific regions and time periods, such as viewing customer flow and sales links in a business district during holiday promotions.

[0130] Secondly, an improved PageRank algorithm can be used to assess the importance of nodes in the graph. Simultaneously, the DBSCAN density clustering algorithm can be applied to cluster graph nodes based on their importance scores, automatically identifying high-density node areas as key marketing touchpoints, such as high-potential promotional zones or clusters of popular products. These touchpoints are visually marked on the graph in the form of a heatmap, with color intensity reflecting the degree of node clustering and commercial value.

[0131] Finally, an interactive analytics interface can be set up, allowing users to click on key marketing touchpoints in the visualization interface to instantly retrieve marketing decision suggestions generated based on graph topology and clustering results. These suggestions cover resource allocation strategies, competitive landscape analysis, and other content, realizing a closed loop from data visualization to intelligent decision-making, significantly improving the application value and decision-making efficiency of industry graphs in scenario-based marketing.

[0132] like Figure 2The diagram shown is a system block diagram of an industry map construction system based on scenario-based marketing provided by an embodiment of the present invention. The construction system includes:

[0133] The data acquisition module is used to acquire social media text through distributed crawlers, process it with a BERT pre-trained model to generate a marketing text dataset with semantic labels, and collect offline scene data through IoT devices, process it with a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal labels. The marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels are combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated.

[0134] The data processing module is used to preprocess the multimodal marketing scenario dataset and construct a four-dimensional dynamic data lake using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology.

[0135] The model building module is used to acquire historical marketing scenario datasets and build a three-dimensional marketing scenario feature library. The marketing scenario weight model is obtained by training the three-dimensional marketing scenario feature library through an LSTM neural network.

[0136] The graph construction module is used to map the four-dimensional dynamic data lake to the Neo4j graph database to form an industry chain topology. Based on the marketing scenario weight model, the strength of the node relationships in the industry chain topology is adjusted. A graph neural network is introduced and the implicit features of the nodes in the industry chain topology are learned through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology.

[0137] Figure 2 The apparatus of the illustrated embodiment can be used to perform corresponding actions. Figure 1 The steps in the method embodiments shown are implemented in a similar manner and have similar technical effects, and will not be repeated here.

[0138] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor performs the steps of an industry map construction method based on scenario-based marketing as described above.

[0139] like Figure 3 The diagram shown is a hardware structure schematic of an electronic device according to an embodiment of the present invention. The electronic device 30 includes: a processor 31, a memory 32, and a computer program; wherein...

[0140] The memory 32 is used to store the computer program, and the memory may also be flash memory. The computer program is, for example, an application program or functional module that implements the above method.

[0141] Processor 31 is configured to execute the computer program stored in the memory to implement the various steps performed by the device in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0142] Alternatively, the memory 32 can be either standalone or integrated with the processor 31.

[0143] When the memory 32 is a device independent of the processor 31, the device may further include:

[0144] Bus 33 is used to connect the memory 32 and the processor 31.

[0145] A readable storage medium storing a computer program, which, when executed by a processor, is used to implement the steps of an industry map construction method based on scenario-based marketing as described above.

[0146] The readable storage medium can be a computer storage medium or a communication medium. A communication medium includes any medium that facilitates the transfer of computer programs from one location to another. A computer storage medium can be any available medium accessible to a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application-Specific Integrated Circuit (ASIC). Alternatively, the ASIC can be located in a user equipment. Of course, the processor and the readable storage medium can also exist as discrete components in a communication device. The readable storage medium can be a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0147] The present invention also provides a program product including executable instructions stored in a readable storage medium. At least one processor of the device can read the executable instructions from the readable storage medium, and the at least one processor executes the executable instructions to cause the device to implement the methods provided in the various embodiments described above.

[0148] In the embodiments of the above-described device, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0149] Through the above embodiments, this invention, through a method and system for constructing an industry graph based on scenario-based marketing, acquires social media text through distributed crawling, processes it using a BERT pre-trained model to generate a marketing text dataset with semantic tags, and simultaneously collects offline scenario data through IoT devices, processes it using a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal tags. The semantically labeled marketing text dataset and the spatiotemporally labeled marketing behavior dataset are then combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated. The Apache Flink stream-batch processing framework and knowledge graph entity disambiguation technology are used to preprocess and construct the multimodal marketing scenario dataset. A four-dimensional dynamic data lake is constructed; historical marketing scenario datasets are acquired and a three-dimensional marketing scenario feature library is built. The marketing scenario weight model is obtained by training the three-dimensional marketing scenario feature library through an LSTM neural network; the four-dimensional dynamic data lake is mapped to the Neo4j graph database to form an industry chain topology. The strength of the node relationship in the industry chain topology is adjusted based on the marketing scenario weight model. A graph neural network is introduced and the implicit features of the nodes in the industry chain topology are learned through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology. This allows the construction of a scenario-sensitive industry graph to reflect the relationship between industry entities under different marketing scenarios in real time, thereby improving the effectiveness and accuracy of the industry graph in marketing decisions.

[0150] This invention integrates semantic data from social media texts and spatiotemporal behavioral data collected from IoT devices, combined with structured data from ERP systems, to generate a multimodal marketing scenario dataset, enabling the construction of industry graphs based on a more comprehensive and richer data foundation. The method utilizes the Apache Flink streaming-batch processing framework to achieve real-time processing of unstructured data and micro-batch processing of structured data, ensuring the timeliness and efficiency of data processing. By constructing a 3D marketing scenario feature library and training a marketing scenario weight model using an LSTM neural network, this invention can adjust the strength of node relationships in the industry chain topology in real time according to different marketing scenarios, making the industry graph scenario-sensitive and more accurately reflecting the relationships between industry entities under different marketing scenarios. This invention combines graph neural networks and the GraphSAGE algorithm to learn the implicit features of nodes, and optimizes the industry chain topology through similarity functions and community detection algorithms, generating a more accurate scenario-sensitive industry graph. Furthermore, this invention provides users with intuitive and effective marketing decision support through a 3D visualization engine, node importance calculation, key marketing touchpoint identification, and interactive analysis interfaces, improving the scientific nature and accuracy of marketing decisions.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing an industry map based on scenario-based marketing, characterized in that, The construction method includes: Social media text is obtained through distributed crawling, and after being processed by a BERT pre-trained model, a marketing text dataset with semantic labels is generated. At the same time, offline scene data is collected through IoT devices, and after being processed by a spatiotemporal calibration algorithm, a marketing behavior dataset with spatiotemporal labels is generated. The marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels are combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated. The multimodal marketing scenario dataset is preprocessed using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology to construct a four-dimensional dynamic data lake. The four-dimensional dynamic data lake refers to a distributed data storage architecture built according to four dimensions: entity, relationship, spatiotemporal, and attribute, which supports real-time incremental data updates and data lineage tracing. Acquire historical marketing scenario datasets and construct a three-dimensional marketing scenario feature library. The three-dimensional marketing scenario feature library refers to a feature set constructed based on historical marketing data, which includes time dimension, spatial dimension and behavioral dimension. It reflects the core attributes of different marketing scenarios through three-dimensional feature vectors. The marketing scenario weight model is obtained by training the three-dimensional marketing scenario feature library through LSTM neural network. The four-dimensional dynamic data lake is mapped to the Neo4j graph database to form an industry chain topology. The strength of node relationships in the industry chain topology is adjusted based on the marketing scenario weight model. A graph neural network is introduced and the implicit features of nodes in the industry chain topology are learned through the GraphSAGE algorithm to generate a scenario-sensitive industry graph topology. The implicit features of nodes refer to the abstract feature vectors containing deep semantic information generated after multi-layer aggregation learning of node features in the industry chain topology through graph neural networks. The scenario-sensitive industry graph topology refers to the industry chain graph used to dynamically adjust the strength of node relationships and topological connections according to the current marketing scenario. It can adaptively present key nodes and core associations that are highly relevant to the marketing scenario. Based on the adjusted edge weights and the learned latent features of nodes, the industry chain topology is recalculated and optimized. First, the latent features of nodes and edge weights are fused by a similarity function to measure the semantic association strength and scene adaptability between nodes, so as to reconstruct the node connection relationship. Second, a community detection algorithm is used to modularize the new topology. By maximizing the modularity index, highly cohesive and loosely coupled node communities are identified. Finally, based on the weighted degree and average similarity of nodes within a community, the importance score of each community is calculated, and the communities with the highest scores are selected as key node clusters to form a scene-sensitive industry graph topology.

2. The method for constructing an industry map based on scenario-based marketing according to claim 1, characterized in that, The process involves acquiring social media text through distributed web crawlers, processing it using a BERT pre-trained model to generate a marketing text dataset with semantic tags, and simultaneously collecting offline scenario data through IoT devices, processing it using a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal tags. The semantically tagged marketing text dataset and the spatiotemporally tagged marketing behavior dataset are then combined to obtain an unstructured marketing dataset. Finally, the structured marketing dataset is combined to generate a multimodal marketing scenario dataset, specifically including: The social media text is obtained in real time through the distributed crawler, and the BERT pre-trained model is used to perform semantic encoding on the social media text. The industry chain entities in the social media text are extracted through named entity recognition technology, and the relationship between the industry chain entities is constructed by dependency parsing. The industry chain entities and their relationships are summarized to generate the marketing text dataset with semantic tags. Simultaneously, the IoT device is connected to the edge computing gateway to collect the offline scene data. The offline scene data is then geofence matched and timestamp calibrated using the spatiotemporal calibration algorithm. This involves mapping the offline scene data to spatial entity coordinates and labeling the collection time granularity of the offline scene data, thus forming the marketing behavior dataset with spatiotemporal tags. The unstructured marketing dataset is obtained by summing the marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels, and the structured marketing dataset is obtained by synchronously accessing the ERP system. The unstructured marketing dataset and the structured marketing dataset are combined to form the multimodal marketing scenario dataset.

3. The method for constructing an industry map based on scenario-based marketing according to claim 1, characterized in that, The process of preprocessing the multimodal marketing scenario dataset and constructing a four-dimensional dynamic data lake using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology specifically includes: The Apache Flink streaming and batch processing framework is used to perform real-time event processing on the unstructured marketing dataset, while micro-batch processing is performed on the structured marketing dataset. The knowledge graph entity disambiguation technology is used to construct an industry chain entity dictionary based on the knowledge graph. A deep semantic matching algorithm is then used to disambiguate the unstructured marketing dataset and the structured marketing data based on the industry chain entity dictionary and to establish a unified entity identifier. The multimodal marketing scenario dataset is mapped to the four-dimensional dynamic data lake based on a distributed storage architecture.

4. The method for constructing an industry map based on scenario-based marketing according to claim 1, characterized in that, The process of acquiring historical marketing scenario datasets and constructing a 3D marketing scenario feature library, and then training the 3D marketing scenario feature library using an LSTM neural network to obtain a marketing scenario weight model, specifically includes: Obtain the historical marketing scenario dataset, and extract the time marketing features, spatial marketing features, and behavioral marketing features from the historical marketing scenario dataset to construct the three-dimensional marketing scenario feature library; The LSTM neural network is used to train the 3D marketing scene feature library, and the marketing weight vectors corresponding to the marketing scene features in the 3D marketing scene feature library are matched to generate the marketing scene weight model.

5. The method for constructing an industry map based on scenario-based marketing according to claim 1, characterized in that, The process involves mapping the four-dimensional dynamic data lake to the Neo4j graph database to form an industry chain topology. Based on the marketing scenario weight model, the strength of node relationships in the industry chain topology is adjusted. A graph neural network is introduced, and the GraphSAGE algorithm is used to learn the implicit features of nodes in the industry chain topology, generating a scenario-sensitive industry graph topology. Specifically, this includes: The four-dimensional dynamic data lake is mapped to the Neo4j graph database to construct the industry chain topology. The enterprise ID, product ID, and user ID in the four-dimensional dynamic data lake are mapped to nodes in the industry chain topology. The supply and demand relationship and public opinion relationship in the four-dimensional dynamic data lake are mapped to edges in the industry chain topology. The spatiotemporal attributes in the four-dimensional dynamic data lake are used as labels for nodes or edges in the industry chain topology. The edge weights, i.e. the node relationship strength, are adjusted in real time based on the marketing scenario weight model. The graph neural network is introduced, and the GraphSAGE algorithm is used to learn the embedded representation of the nodes. The features of adjacent nodes are weighted and aggregated in combination with the node relationship strength to obtain the hidden features of the nodes. The industry chain topology is recalculated based on the node relationship strength and the node implicit features, and the key node clusters in the updated industry chain topology are identified by the community discovery algorithm to generate the scenario-sensitive industry graph topology.

6. The method for constructing an industry map based on scenario-based marketing according to claim 5, characterized in that, The real-time adjustment of edge weights based on the marketing scenario weight model specifically includes: The FastText text classification model is used to extract keywords and perform semantic classification on the current marketing scenario description text in the four-dimensional dynamic data lake to determine the corresponding current marketing scenario type and extract the corresponding current marketing scenario features. Based on the marketing scenario weight model, the current marketing weight vector corresponding to the current marketing scenario feature is matched, which is the edge weight.

7. The method for constructing an industry map based on scenario-based marketing according to claim 5, characterized in that, The entity set in the four-dimensional dynamic data lake is defined as follows: ,in Let i represent the i-th entity, i.e., enterprise ID, product ID, or user ID, and n represent the total number of entities in the four-dimensional dynamic data lake. The entity set E is mapped to the node set in the industry chain topology G. ,and And map() represents the mapping function from entity to node; the set of edges in the industry chain topology G is defined as ,in Represents the i-th node and the j-th node The relationship types are defined as follows: the spatiotemporal attribute set in the four-dimensional dynamic data lake is... ,in Indicates time attribute, Indicates spatial attributes; The formula for calculating the hidden features of the nodes is as follows: In the formula, Represents a node The hidden features of nodes in the l-th layer of a graph neural network; Indicates the activation function; This represents the feature aggregation function of the l-th layer; Represents a node Hidden features of nodes in the (l-1)th layer of a graph neural network; This represents the adjusted edge weight, i.e., the node weight. and The strength of the node relationships between them; Represents a node The set of neighboring nodes; Represents a node Hidden features of nodes in layer l-1 of a graph neural network.

8. The method for constructing an industry map based on scenario-based marketing according to claim 7, characterized in that, The step of recalculating the industry chain topology based on the node relationship strength and the node implicit characteristics specifically includes: Based on the node relationship strength and the node implicit features, the similarity function is constructed as follows: In the formula, Represents a node and nodes The similarity function; L represents the last layer of the graph neural network; Represents a node Hidden features of nodes in the last layer of a graph neural network; Represents a node Hidden features of nodes in the last layer of a graph neural network; Representing the hidden features of nodes The L2 norm; Representing the hidden features of nodes The L2 norm; Represents a node Hidden features of nodes With nodes Hidden features of nodes Cosine similarity; The empirical coefficient representing the amplification factor of the node relationship strength. This represents the amplification factor for the strength of the node relationship. ; Set similarity threshold ,when At that time, retain the edge Otherwise delete the edge. The industry chain topology is recalculated to obtain the updated industry chain topology. ; The step of identifying key node clusters in the updated industry chain topology using a community discovery algorithm to generate the scenario-sensitive industry graph topology specifically includes: The updated industry chain topology Each node in the process forms a community. ,in Indicates the i-th community; The community detection algorithm is used to calculate the communities respectively. and neighboring communities The community module level is and and calculate the community With neighboring communities The modularity of the merged new community is The community modularity gain is obtained as ; If the community modularity gain Then merge the communities. and neighboring communities Otherwise, maintain the community. and neighboring communities independent; The final set of communities is obtained by iterating through all adjacent community pairs. , where p represents the final community set The total number of communities in the area; Calculate the final community set Each community Community Importance Score ,in Importance score for all communities Arrange the communities from largest to smallest, select the top M communities corresponding to their importance scores to form a key community cluster, i.e., the key node cluster, and generate the scenario-sensitive industry map topology structure.

9. The method for constructing an industry map based on scenario-based marketing according to claim 1, characterized in that, It also includes setting up a WebGL-based 3D visualization engine that allows users to perform spatiotemporal slicing of the scene-sensitive industry map topology using geofences and timestamps. An improved PageRank algorithm is used to calculate the importance of nodes in the scenario-sensitive industry graph topology, and the DBSCAN clustering algorithm is used to identify key marketing touchpoints based on the importance of the nodes and generate marketing decision suggestions. The key marketing touchpoints are marked in the form of a heat map. An interactive analytics interface is set up to allow users to click on the key marketing touchpoints and view the marketing decision recommendations.

10. An industry map construction system based on scenario-based marketing, applied to the industry map construction method based on scenario-based marketing as described in any one of claims 1-9, characterized in that, The construction system includes: The data acquisition module is used to acquire social media text through distributed crawlers, process it with a BERT pre-trained model to generate a marketing text dataset with semantic labels, and collect offline scene data through IoT devices, process it with a spatiotemporal calibration algorithm to generate a marketing behavior dataset with spatiotemporal labels. The marketing text dataset with semantic labels and the marketing behavior dataset with spatiotemporal labels are combined to obtain an unstructured marketing dataset. Combined with the structured marketing dataset, a multimodal marketing scenario dataset is generated. The data processing module is used to preprocess the multimodal marketing scenario dataset and build a four-dimensional dynamic data lake using the Apache Flink stream and batch processing framework and knowledge graph entity disambiguation technology. The four-dimensional dynamic data lake refers to a distributed data storage architecture built according to four dimensions: entity, relationship, spatiotemporal, and attribute, which supports real-time incremental data updates and data lineage tracing. The model building module is used to acquire historical marketing scenario datasets and build a three-dimensional marketing scenario feature library. The three-dimensional marketing scenario feature library refers to a feature set built based on historical marketing data, which includes time, space and behavior dimensions. It reflects the core attributes of different marketing scenarios through three-dimensional feature vectors. The marketing scenario weight model is obtained by training the three-dimensional marketing scenario feature library through LSTM neural network. The graph construction module is used to map the four-dimensional dynamic data lake to the Neo4j graph database to form an industry chain topology. Based on the marketing scenario weight model, the strength of node relationships in the industry chain topology is adjusted. A graph neural network is introduced and the GraphSAGE algorithm is used to learn the implicit features of nodes in the industry chain topology to generate a scenario-sensitive industry graph topology. The implicit features of nodes refer to the abstract feature vectors containing deep semantic information generated after multi-layer aggregation learning of node features in the industry chain topology through graph neural network. The scenario-sensitive industry graph topology refers to the industry chain graph used to dynamically adjust the strength of node relationships and topological connections according to the current marketing scenario. It can adaptively present key nodes and core associations that are highly relevant to the marketing scenario. Based on the adjusted edge weights and the learned latent features of nodes, the industry chain topology is recalculated and optimized. First, the latent features of nodes and edge weights are fused by a similarity function to measure the semantic association strength and scene adaptability between nodes, so as to reconstruct the node connection relationship. Second, a community detection algorithm is used to modularize the new topology. By maximizing the modularity index, highly cohesive and loosely coupled node communities are identified. Finally, based on the weighted degree and average similarity of nodes within a community, the importance score of each community is calculated, and the communities with the highest scores are selected as key node clusters to form a scene-sensitive industry graph topology.

Images